The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
STUDIESON
CAUSES AND
CONSEQUENCES
1989
LOWDOWN
Federal Reserve Bank of New York
Studies on
Causes and Consequences of the
1989-92 Credit Slowdown
Federal Reserve Bank of New York
February 1994
Contents
Forward
by William J. McDonough
Causes and Consequences of the 1989-92 Credit Slowdown:
Overview and Perspective
by M. A. Akhtar
i
1
Economic Activity and the Recent Slowdown in Private Sector Borrowing
by Patricia C. Mosser and Charles Steindel
39
The Role of the Banking System in the Credit Slowdown
by Cara Lawn and John Wenninger
69
The Link Between the 1980s Credit Boom and
the Recent Bank Credit Slowdown
by Ronald Johnson and Chun K. Lee
113
Loan Sales and the Slowdown in Bank Lending
by Rebecca Demsetz
131
Foreign Credit Expansion in the United States
by Rama Seth
149
Nonbank Lenders and the Credit Slowdown
by Richard Cantor and Anthony P. Rodrigues
171
Survey Evidence on Credit Tightening and
the Factors Behind the Recent Credit Crunch
by Kausar Hamdani, Anthony P. Rodrigues, and Maria Varvatsoulis
211
Influence of the Credit Crunch on Aggregate Demand and
Implications for Monetary Policy
by Patricia C Mosser
259
The Credit Crunch and the Construction Industry
by Ethan S. Harris, Michael Boldin, and Mark D. Flaherty
301
Credit Supply Constraints on Business Activity, Excluding Construction
by Charles Steindel and David Brauer
355
The Credit Slowdown and the Monetary Aggregates
by R.S. Hilton and C.S. Lown
397
The Credit Slowdown Abroad
by 5. Hickok and C Osier
429
FOREWORD
The 1990-91 recession and the recent economic expansion have been
marked by an exceptional weakness in bank and nonbank credit flows. A major cause of the credit weakness has been tepid loan demand stemming from
the general sluggishness of economic activity in recent years. That does not
appear to be the whole story, however. The need to reduce historically unprecedented business and household debt burdens that had resulted from
heavy debt accumulation in the 1980s has been an important element underlying the recent credit slowdown. At the same time, for a variety of reasons,
both bank and nonbank lenders appeared to have restrained credit supply significantly over 1989-92. The confluence of these credit market problems
clearly played a role in the disappointing performance of the economy in recent years.
Against this background, it seemed worthwhile to us at the Federal Reserve Bank of New York to examine a broad range of important issues concerning the recent credit slowdown. The results of that examination are
contained in the present volume. The papers deal with, among other issues,
the importance of credit demand versus credit supply factors, the role of bank
and nonbank credit sources, the impact of credit supply shifts on the economy
and the implications of those shifts for monetary policy. The first paper in the
collection contains a summary of the main findings and a perspective on the
implications for monetary policy.
Overall, this collection provides a comprehensive analysis of the 1989-92
credit slowdown experience, and should make a significant contribution to
understanding the recent problems in the nation's credit markets. We also
hope that the results and insights of the study will prove useful to all those
concerned with issues of monetary policy.
William J. McDonough
February 1994
Causes and Consequences of the 1989-92 Credit
Slowdown: Overview and Perspective
by M. A. Akhtar1
Between early 1989 and late 1992, U.S. economic growth averaged less than 1 percent,
well below the long-run trend growth of the economy. This sluggish pattern of growth
persisted in the face of substantial easing in monetary policy. Indeed, the economy
failed to recover significantly after the 1990-91 downturn. Apparently, the favorable effects of monetary easing were not sufficient to overcome numerous factors depressing
the economy: lower defense spending, commercial real estate depression, relatively
tight fiscal policy, global competition, corporate restructuring, historically low levels of
consumer confidence, and the overextended financial positions of households, businesses, and financial institutions.
The sluggish real growth was accompanied by an unprecedentedly sharp slowdown
in credit growth over 1989-92. Many observers have identified high debt service burdens of the nonfinancial sectors and widespread balance sheet problems of borrowers
and lenders as crucial elements underlying both the credit slowdown and the persistent
weakness of the economy. Others have attributed the sluggish economic performance
to supply-side factors underlying the credit slowdown, which resulted in a prolonged
period of substantially reduced credit availability to businesses and households. More
recently, concerns about credit availability appear to have eased as credit growth has
shown some signs of recovery.
Against the background of these developments, this overview provides a broad perspective on the causes and consequences of the 1989-92 credit slowdown. It begins by
presenting a general conceptual framework for the analysis and then reviews the evidence from the collection of studies on the credit slowdown. The article also discusses
implications of the evidence for monetary policy and offers some tentative general observations on the recent credit slowdown experience.
Overall, studies reviewed here provide substantial evidence of credit supply problems, or a "credit crunch" during the 1989-92 period for both bank and nonbank credit
sources. The evidence on the consequences of credit supply constraints is less compel1
I am grateful to Richard Davis and Glenn Hubbard for comments and discussions, and to Martina Hcyd for
competent research assistance.
Causes and Consequences
ling, but the studies do indicate, at least collectively, that credit constraints have played
some role in weakening economic activity. The depressing effects of the credit crunch
appear not to have been the primary or dominant cause of the economic slowdown, however. As for the implications for monetary policy, credit supply problems clearly contributed to reducing the effectiveness of monetary policy, although it is difficult to
isolate their effects from those of other factors disrupting or altering the channels of policy influence to the economy.
I. Credit Slowdown vs. Credit Crunch: A General Framework
There is no generally accepted definition of the term "credit crunch," but it is usually
taken to mean a sharp reduction in the supply or availability of credit at any given level
of interest rates. To clarify terminology and to provide a broad context for the issues
involved in identifying a credit crunch, we begin with the more encompassing notion of
credit slowdown or decline. At the broadest level, an observed slowdown or decline in
credit may result from either the demand side or the supply side. At a given lending rate
or the price of credit, the demand for credit may fall because of other (nonprice) determinants of credit demand. In the usual graphical supply-demand framework, the demand schedule for credit may shift down and to the left. This is shown in Chart 1, panel
I, under very simplistic market conditions, where the price of credit includes both the
loan rate and nonrate loan terms, such as collateral, maturity, and covenants. From a
macroeconomic perspective, this type of shift may occur because of lower credit demand stemming from either cyclical weakness in economic activity or structural factors—such as changes in the tax code, inventory techniques, or the borrowers' desired
debt-to-income ratio—that reduce the perceived need for credit permanently. In general, shifts in credit demand induced by cyclical weakness in economic activity are relatively commonplace while credit demand shifts due to structural changes are somewhat
less frequent but not unusual.
A downward shift in credit demand tends to put downward pressures on loan rates
and other loan terms and, given an unchanged supply schedule, leads to easier loan
terms at the new credit market equilibrium. Moreover, if a downward credit demand
shift is caused by structural factors, it may also be accompanied by a steepening (flattening) of the demand schedule; the demand for credit may become less (more) responsive to changes in the price of credit (Chart I, panel I, D2 schedule).
On the supply side, a credit slowdown or decline may reflect reduced willingness to lend
at prevailing interest rates and demand conditions. Factors that can cause reduced willingness to lend include, among others, balance sheet difficulties of lenders (poor quality
assets, high loan losses, etc.), higher capital requirements and regulatory constraints on
lenders, and increases in actual or perceived riskiness of borrowers' credit quality. The
last factor is intended to capture credit supply shifts resulting from changes in a borrower's balance sheet conditions. Specifically, a deterioration in the quality of a borrower's
balance sheet reflecting, for example, a drop in asset prices, weakens his ability to repay
existing debts or to borrow new funds.2 The decline in creditworthiness of the borrower,
in turn, may reduce the lender's willingness to extend a loan, causing a decline in the
supply of credit. In this situation, the supply shift reflects reduced credit availability to
2
More generally, the deterioration in the quality of the borrower's balance sheet (and the associated decline
in the credit wort hi ness) may result either from a cyclical decline, or from noncyclical shocks (economywide or partial) such as an asset price drop in one or more sectors. As discussed below, it is very difficult
to separate credit supply from demand effects of general cyclical shocks to the economy.
desire to lend to those borrowers whose creditworthiness has remained unchanged.
Note that, the drop in borrowers' creditworthiness could be treated, in principle, as a
drop in credit demand by borrowers of given risk characteristics (unchanged creditwor-
Chart 1: Credit Demand and Supply
Price of Credit
Panel 1
Panel 2
D = initial credit demand schedule, where all determinants_other than the price of credit
(loan rate and nonrate loan terms) are held constant (y is a proxy for aggregate
demand in the economy and x represents all other variables).
S = initial credit supply schedule, where alljdeterminants other than the price of credit are
held constantjk is a proxy for capital, g is a proxy for regulatory and supervisory constraints, and z represents all other variables).
C*= initial equilibrium credit.
r* = initial equilibrium price of credit.
DT and D 2 represent the credit demand schedule after shifts, while ST and S 2 represent the
credit supply schedule after shifts. C*1f C*2, r*1f and r*2 represent new equilibrium
values for credit and the price of credit.
Causes and Consequences
borrowers whose credit quality has been impaired, but there is no change in the lender's
thiness) in that there are fewer such borrowers. Nonetheless, at a practical level, it is
more convenient to look at the effect of changes in borrowers1 credit quality—especially
those resulting from noncyclical shocks—on the willingness of lenders to supply credit.
In any event, the reduced willingness to lend may show up as a leftward shift in the
credit supply schedule (Chart I, panel 2). In this case, borrowing is rationed by price as
loan rates and nonrate loan terms tend to tighten and the new credit market equilibrium
is attained at higher interest rates and generally more restrictive loan terms, other things
equal.
The reduced willingness to lend may not show up as a simple leftward shift of credit
supply envisaged in the context of a market-clearing environment, however. Instead,
lenders may resort to increased nonprice credit rationing, that is, loans are rationed by
quantity rather than by variations in prices (interest rates and nonrate loan terms). In
this case, lenders do not feel that they can protect themselves against risk by charging
higher credit prices. Put another way, the credit supply schedule is not fully operative;
in the extreme case, the schedule shifts leftward and becomes vertical, with the supply
of credit becoming completely insensitive to interest rates (Chart 1, panel 2, S2 schedule). In practice, the existence of nonprice credit rationing does not preclude the role of
interest rates and other loan terms; some borrowings may be rationed by price and others
by quantity or by both. Nonprice credit rationing may take many different forms: some
borrowers obtain loans while other borrowers with identical creditworthiness do not;
loans for certain types of borrowing or to certain classes of borrowers are unavailable;
some apparently creditworthy borrowers are denied loans at prevailing interest rates because lenders1 do not perceive them to be creditworthy.3
The papers in this volume deal with both demand and supply factors in the credit
slowdown since 1989, but the emphasis is on sorting out the role of supply-side factors
and their implications for nonfinancial economic activity. Accordingly, the term "credit
crunch" as used here refers to a slowdown or decline in the supply of credit, whether
rationed by price or nonprice mechanisms, or simply to credit supply problems. This
definition is clearly much broader than the narrow use of that term to describe situations
of nonprice credit rationing. It is also broader than another frequently mentioned definition of credit crunch: "a widespread, sudden, sharp, indiscriminate, and rather brief
credit shutdown" (Wojnilower, I993). 4
In a macroeconomic context, the existence of credit supply problems implies that the
observed credit slowdown (reduction) cannot be fully explained by cyclical developments in aggregate demand, except insofar as cyclical developments may have significant adverse effects on borrowers1 creditworthiness as perceived by lenders. There are,
of course, numerous identification problems in sorting out supply from demand factors
in the credit slowdown. For example, a sharp reduction in the willingness to lend may
lead to a decline in output, inducing a reduction in the demand for credit. In these circumstances, the credit slowdown will be reported as reflecting lower demand for credit
even though it was, in fact, caused by an initial shock to the supply of credit (Friedman
1993 a, b).
•* Sec JaiTec and Slight/. (1990) For a detailed survey of various aspects of credit rationing.
For other perspectives on defining a credit crunch, sec Peek and Rosengren (1992), Owens and Schreft
(1992) and Wojnilower (1992a). For other perspectives on the current credit crunch, see Bcrnankc and
Lown (1991), Cantor and Wenninger (1993), Jones (1993), Jordan (1992), Kaufman (1991). Kliescn and
Tatom (1992), Peek and Rosengren (1992), Sinai (1993), Syron (1991), and Wojnilower (1993). For
detailed analysis of earlier crunches, sec Wojnilowcr (1980) and Wolfson (1986).
More generally, with both demand and supply factors operating simultaneously and
interacting with each other, it is very difficult to distinguish shifts in the supply schedule
from developments on the demand side. Lenders usually tend to tighten credit standards
and terms for lending when the overall economy slips into a recession because, on average, business and household loans entail higher risks than before. But the extent of
lenders' response depends not only on the degree of perceived economic weakness and
its effects on borrowers' credit quality but also on the state of their own balance sheets.
From the perspective of borrowers, this situation would look like a contraction in credit
supply while lenders may believe this to be a response to developments in aggregate demand. Strictly speaking, there is no change in the lenders' willingness to extend credit
to borrowers of given circumstances (that is, unchanged creditworthiness). At the same
time, the reduced supply is not a response to lower demand for credit. The constriction
in the supply of credit has clearly been caused by a decline in the willingness of lenders,
albeit reflecting the adverse effect of the weaker economy on the creditworthiness of
borrowers and balance sheets of banks. Any sorting out of the demand and supply aspects in this case would be further complicated by the fact that the recession itself would
reduce the demand for credit.
The identification of demand and supply factors in the recent credit slowdown is particularly difficult because of the conjunction of the prolonged cyclical weakness in the
economy with a correction of earlier credit excesses. Those credit excesses, as noted in
Section HID below, reflected the process of unusually rapid increases in debt in the mid1980s and became unsustainable over time as both borrowers and lenders experienced
balance sheet and other difficulties, with cyclical developments reinforcing pressures
for correction. In this highly "endogenous" process, the demand for credit is believed
to have fallen simultaneously with reductions in banks' capacity and willingness to lend.
Notwithstanding these difficulties, the twelve studies in this volume examine a broad
range of issues concerning the 1989-92 credit slowdown. Five of these studies (Lown/
Wenninger, Cantor/Rodrigues, Johnson/Lee, Demsetz, Seth) look at various aspects of
the role of bank and nonbank credit sources in the slowdown of private nonfinancial
debt, focusing on the importance of credit demand relative to credit supply factors. One
study (Hamdani/Rodrigues/Varvatsoulis) reviews survey data on credit tightening from
lenders and borrowers, and another study (Mosser/Steindel) explores the role of economic activity and other "fundamentals" in explaining the recent credit slowdown.
Three studies (Harris/Boldin/Flahcrty, Mosser, Steindel/Brauer) investigate the effects
of credit supply problems on various aspects of nonfinancial economic activity. Finally,
two studies (Hilton/Lown, Hickok/Osler) consider some special aspects of the credit
slowdown: one attempts to assess the impact of credit supply shifts on the broadly defined money stock, M 2 , and the other provides a broad overview of the nature and extent
of the credit slowdown abroad, largely based on the experience in France, Japan, and the
United Kingdom.
The remainder of this article reviews evidence from the twelve studies under four
broad headings: the extent of the credit slowdown; factors behind the credit slowdown;
consequences of the credit crunch for nonfinancial economic activity; and implications
of the credit crunch for monetary policy. The last section offers a few tentative concluding observations on the recent credit crunch experience.
II. Extent of the Recent Credit Slowdown
Collectively, the studies in this volume show that the U.S. economy has experienced a
broadly based and sharp credit slowdown in recent years. In documenting and describ-
Causes and Consequences
ing the credit slowdown from the viewpoint of various types of borrowers (business,
household, real estate, small business) or lenders (banks, other depositories, finance
companies, insurance companies, foreign banks, bond markets), most of the studies begin by examining the extent of credit slowdown in the recent period. Since the timing
of the slowdown is not uniform across all borrowers and lenders, however, these studies
do not target a common time period for the recent credit slowdown. Nor do they judge
the recent credit slowdown against a common historical benchmark. Instead, each study
provides a comprehensive look at relevant credit developments from its particular vantage point using whatever time periods make most sense.
Nevertheless, it may be useful to provide a common time frame for summarizing the
extent of the slowdown in private nonfinancial debt and its main components on both
the lending and the borrowing sides. I use the flow of funds data to highlight the breadth
and depth of credit slowdown over the last three years (1989-1V to 1992-1V), taken as a
whole, relative to long-term trends in the periods 1960-82 and 1982-89. Because inflation was greater in the earlier periods than in the most recent period, comparisons of
nominal credit growth rates may be misleading. I have, therefore, presented data in both
nominal and real terms in many cases. For simplicity and convenience, however, I have
used the GDP deflator to convert nominal dollars into real dollars rather than search for
specific sectoral deflators. (Sectoral deflators might change precise real dollar values
but they are unlikely to alter the broader contours of constant dollar data obtained on the
basis of the GDP deflator.) The points made here provide a broad overview of the extent
of the credit slowdown to nonfinancial borrowers from both bank and nonbank sources,
and may be viewed as a summary of details in various studies.
A. Private Nonfinancial Debt
Using data on nominal and real debt and ratios of debt to GDP, I begin by looking at the
extent of the slowdown in private nonfinancial debt in terms of its three broad decompositions: business versus household debt, mortgage versus nonmortgage debt, and corporate versus noncorporate debt. As shown in Table 1, private nonfinancial debt growth
declined sharply to about 3 percent, at an annual rate, over 1989-92 from long-term
trend rates of 9 1/2-10 1/2 percent. Both businesses and households experienced large
debt slowdowns, but the rate of decline was much greater for the business sector. Nonfinancial business sector debt growth averaged less than 1 percent in the recent period,
compared with a long-term trend rate of 10 percent, while household debt growth averaged 5.6 percent in the recent period, about one-half the average growth rate over 198289.
In real terms, private nonfinancial debt actually declined somewhat over 1989-92
compared with trend rates of nearly 7 percent and 4 1/4 percent over 1982-89 and 196082, respectively. For both the business and household sectors, real debt trend growth
rates were significantly higher in the 1982-89 period than in the earlier period. Credit
to the nonfinancial business sector declined by nearly 3 percent, on average, in real
terms over 1989-92, following more than 6 percent average growth over 1982-89. The
sharp declines in private and business debt growth in recent years have reversed the rising trends of ratios of private and business sector debt to GDP (Chart 2 and Table 1) despite a sustained period of weak growth of nominal GDP.
With nonmortgage debt of both businesses and households slowing to about 2 percent at an annual rate over 1989-92, the greater decline in total business debt growth relative to household debt growth in recent years appears to be largely the result of
differences in home and business mortgage debt developments (Table 2). Home mort-
Table 1: Nonfinancial Debt
Fourth Quarter over Fourth Quarter Percent Change, Annual Rate
Total
Nonfinancial
Private
Nonfinancial
Nonfinancial
Business
Households
1960-82
8.6
9.6
10.0
9.2
1982-89
11.0
10.6
10.1
11.1
1989-92
5.2
3.1
0.7
5.6
1960-82
3.1
4.2
4.5
3.8
1982-89
7.2
6.8
6.3
7.3
1989-92
1.7
-0.4
-2.8
2.0
1960-82
0.2
1.2
1.6
0.8
1982-89
3.5
3.1
2.6
3.6
1989-92
0.3
-1.8
-4.2
0.6
100.0
65.3
31.4
33.9
Current dollars
Constant 1987 dollars3
Ratio of debt to GDP
Memo: 1992-IV
current dollar
share of total
nonfinancial debt
a
GDP deflator was used to construct constant dollar series.
gage debt advanced at a hefty 7 percent annual rate in the 1989-92 period, although its
rate of growth decelerated substantially from the historically high average growth rate
over 1982-89. By contrast, business debt for real estate development declined at an average annual rate of about 2 percent during 1989-92, down from an average annual
growth rate of close to 10 percent in the earlier period.
In real terms, both mortgage and nonmortgage components of business debt declined
significantly in the 1989-92 period. But businesses have experienced a much sharper
decline in credit flows for mortgages than for other activity in recent years.
Recent business debt developments have also differed significantly by the size of
borrowers. As a group, large or corporate business borrowers fared better than small or
noncorporate borrowers in the recent credit slowdown. Credit to corporate borrowers
increased at an annual average rate of nearly 2 percent during the last three years, down
from an 11.3 percent average increase over 1982-89 (Table 3). By contrast, noncorporate borrowers experienced an outright credit decline of 1.3 percent, at an annual rate,
in the 1989-92 period compared with growth rates of about 11 percent in 1982-89. It is
interesting to note that noncorporate borrowing is the only category among those reported here that showed significantly lower real debt growth in the 1982-89 period than in
the earlier period.
Causes and
Consequences
Chart 2: Ratios of Debt to GDP
Ratio
1.4
Sources: Board of Governors of the Federal Reserve System, Flow of Funds Accounts;
U.S. Department of Commerce.
Table 2: Private Nonfinancial Debt
Fourth Quarter over Fourth Quarter Percent Change, Annual Rate
Nonmortgage
Mortgage
Private
Nonfinancial
Total
1960-82
9.6
9.6
10.3
1982-89
10.6
10.9
1989-92
3.1
Total
Business
Household
9.3
9.6
9.9
9.1
9.7
11.5
10.3
10.3
10.3
4.2
-1.9
7.1
2.0
1.9
2.3
Home
Business Mortgage
Current dollars
Constant 1987 dollars8
1960-82
4.2
4.2
4.8
3.8
4.2
4.4
3.7
1982-89
6.8
7.1
5.9
7.8
6.5
6.5
6.5
1989-92
-0.4
0.7
-5.4
3.6
-1.4
-1.6
-1.1
100.0
50.2
14.3
35.9
49.8
33.8
16.0
Memo: 1992-1V
current dollar
share of private
nonfinancial debt
a
Based on GDP deflator.
B. Bank and Nonbank Credit Sources
The slowdown in private nonfinancial debt growth was broadly spread across depository
(banks and thrifts) and nondepository credit sources (Table 4). Banks and thrifts, however, experienced a sharper decline in credit growth over 1989-92 than did overall nondepository credit growth. Total depository credit actually declined at an annual rate of
about 2 percent over 1989-92 following 9.3 percent average growth over 1982-89, while
total nondepository credit growth slowed to a 7 percent average rate in the recent period
from about 12 percent in the preceding period. Both depository and nondepository credit growth rates are, of course, much lower on a constant dollar basis. At this level of
aggregation, the bulk of the deceleration in private nonfinancial credit growth over
1989-92 relative to the 1982-89 average rate is accounted for by depository sources,
with both banks and thrifts making substantial contributions to the slowdown.
The outright decline in total depository credit over 1989-92 reflects, to a considerable extent, the collapse of the savings and loan industry. In fact, the commercial bank
credit component—which represents about 70 percent of total depository credit—advanced at a 2 percent average annual rate over the 1989-92 period, compared with a
long-term trend rate of around 10 percent. This modest bank credit growth was more
than fully offset, however, by a 45 percent (13 1/3 percent at an annual rate) decline in
credit by savings and loan associations.
While overall nondepository credit growth has held up better than overall depository
or bank credit growth, many components of nondepository credit did not fare much better than bank credit. As explained in the Cantor/Rodrigues study, credit growth to businesses experienced roughly similar slowdowns in commercial paper, finance company
lending, and bank loans in recent years relative to earlier trends.
Causes and Consequences
Comparing the contribution of depository and nondepository sources to business
credit developments reveals that banks and thrifts accounted for about four-fifths of the
fall in business mortgage debt growth in 1989-92 relative to 1982-89 (Table 5). The
slowdown in nonmortgage business debt in the recent period relative to the earlier period was somewhat more evenly divided between depository and nondepository sources.
For the nonfinancial business sector as a whole, most of the deceleration in the average
credit growth from the 1982-89 period to the 1989-90 period reflected the slowdown in
depository credit; banks accounted for somewhat more than one-half of the depository
contribution.
On the household side, the collapse of the savings and loan industry and the lending
slowdown by other thrifts were responsible for most of the slowdown in home mortgage
debt growth in 1989-92 relative to 1982-89. The pace of commercial bank credit flows
for home mortgages actually picked up somewhat during the 1989-92 period. Banks,
however, made the largest contribution to the slowdown in nonmortgage household
credit, accounting for more than half of the total slowdown in that component.
C. Selected Aspects of Bank Business Loans
Data reported above clearly indicates that commercial banks have played a major role
in the 1989-92 credit slowdown for both business mortgages and nonmortgage business
loans. For the nonfinancial business sector as a whole, the slowdown in bank loans accounted for more than one-third of the deceleration in average credit growth from 198289 to 1989-92.
Table 3: Nonfinancial Business Debt
Fourth Quarter over Fourth Quarter Percent Change, Annual Rate
10
By Size of Borrower
By Type of Borrowing
Total3
Large6
Small0
Mortgage
Other
1960-82
10.0
8.7
14.1
10.3
9.9
1982-89
10.1
11.3
10.9
9.7
10.3
1989-92
0.7
1.8
-1.3
-1.9
1.9
1960-82
4.5
3.3
8.5
4.8
4.4
1982-89
6.3
7.5
7.1
5.9
6.5
1989-92
-2.8
-1.7
-4.8
-5.4
-1.6
Memo: 19 92-IV current dollar share of
private nonfinancial
debt
48.1
31.4
15.0
14.3
33.8
Current dollars
Constant 1987 dollars11
a
- All corporate and noncorporate debt.
Corporate sector, excluding farm debt.
c
- Nonfarm, noncorporate debt.
d
Based on GDP deflator.
b
Both large (corporate) and small (noncorporate) business borrowers from banks experienced outright declines in bank loans over 1989-92, but the rate of decline was considerably greater for noncorporate borrowers (Table 6). Specifically, over the 1989-92
period, nonmortgage bank loans to noncorporate borrowers declined at a 4 1/2 percent
annual rate, more than twice the pace of decline for corporate borrowers.
Table 4: Nonfinancial Private Credit Growth
Fourth Quarter over Fourth Quarter Percent Change, Annual Rate
Depository
Credit
Nondepository Credit
Bank Credit
Depository
Loans
Bank Loans
1960-82
9.7
9.6
10.1
9.7
10.3
1982-89
9.3
11.8
10.1
9.0
9.9
1989-92
-2.0
7.0
2.0
-2.7
1.1
Current dollars
Constant 1987 dollars8
1960-82
4.2
4.1
4.7
4.3
4.8
1982-89
5.5
8.0
6.3
5.3
6.1
1989-92
-5.4
3.5
-1.5
-6.1
-2.4
Memo: 1992IV current dollar share of
private nonfinancial debt
39.9
60.1
28.1
36.3
25.5
a
GDP deflator was used to construct constant dollar series.
Table 5: Contributions to the Credit Slowdown
From 1982-89 to1989-92
Business
Decline in credit
growth ratea
Household
Mortgage
Other
Total
Home
Mortgage
Other
Total
11.6
8.4
9.4
4.4
8.0
5.6
Percent of total decline contributed by:
Depository
sources
82.8
58.3
69.1
84.1
67.5
78.6
Banks
38.8
21.4
37.2
-6.8
55.0
23.2
Thrifts
44.0
36.9
31.9
90.9
12.5
55.4
Nondepository
sources
17.2
41.7
30.9
15.9
32.5
21.4
• Annual average credit growth rate over 1982-89 minus annual average growth rate over 1989-92.
Causes and Consequences
In the absence of bank loan sales, bank credit flows to businesses would probably
have been even weaker in recent years. The study by Demsctz indicates, however, that
adjustments for bank business loan sales to nonbanks and nonfinancial institutions over
the 1986-92 period actually increase the severity of the recent slowdown in commercial
and industrial loans on banks' books because business loan sales have decreased in recent years. (Note that the flow of funds data for nonfinancial borrowers reported here
already incorporate loan sale adjustments.) Even so, the liquidity provided by loan sales
and securitization has most likely enabled banks to maintain higher levels of total loan
origination than would have been the case otherwise. Cantor and Rodrigues point out
in their study for this volume that mortgage-backed securities have grown about 70 percent since 1988 and that securitization of business and consumer credit has proceeded
even more rapidly over that period.^ Clearly, recent sharp advances in securitization
have, to some extent, cushioned the credit slowdown.
As described in detail in the study by Lown and Wenninger, the bank credit slowdown was spread fairly broadly across various regions of the country, but Northeast
(New England and Mid-Atlantic) and Pacific regions experienced very large outright
declines in total and business bank loans over 1989-92. Other regions also experienced
contractions in commercial and industrial loans, although in some cases the rates of decline were relatively modest.
5
Cantor and Demsctz (1993) show that over the two years to the second quarter of 1992, the growth in loans
lor home mortgages, consumers, and businesses inclusive of off-balance-sheet lending (sccuriti/alion and
loan sales) exceeded the growth in loans on the books of banks, thrifts, mortgage companies, and finance
companies as a group.
Table 6: Nonfinancial Business Loans by Banks
Fourth Quarter over Fourth Quarter Percent Change, Annual Rate
Nonmortgage Business Loans
Total
Total3
Large
Business13
Small
Business0
Mortgages
1960-82
10.6
10.3
10.0
14.0
12.0
1982-89
9.9
7.2
8.0
7.1
16.5
1989-92
-1.7
-2.3
-2.2
-4.5
-0.7
1960-82
5.2
4.9
4.5
8.5
6.5
1982-89
6.1
3.5
4.3
3.3
12.7
1989-92
-5.2
-5.7
-5.6
-7.9
-4.2
Memo: 1992-1V
share of private
nonfinancial debt
13.7
8.7
6.9
1.4
5.0
Current dollars
Constant 1987 dollars*
a
All corporate and noncorporate business.
Nonfarm corporate business.
c
- Nonfarm, noncorporate business.
d
- Based on GDP deflator.
b
12
Chart 3: Comparison of Domestic and Foreign Bank Loan Shares
Source: Board of Governors of the Federal Reserve System, Flow of Funds
Note: Shaded areas indicate periods designated recessions by the National
Economic Research.
Within the banking system, the bulk of the recent bank credit slowdown is attributable
to domestic banks as opposed to foreign banking offices in the United States (Chart 3).
Total loans of U.S.-chartered banks showed less than 1 percent annual average growth
over 1989-92, and business loans actually declined outright at a 4.5 percent annual rate.
By contrast, total U.S. loans of foreign banking offices in the U.S. advanced at an annual
rate of about 14 percent over the recent three-year period, only slightly below the average
increase over the 1982-89 period. Business loans by foreign banking offices did register
a significant slowdown in the recent period, but they continued to increase at a hefty annual pace of about 9 percent.
13
Causes and Consequences
These trends in foreign bank loans to U.S. borrowers are analyzed in more detail by
Rama Seth in her study for this collection. She finds that as a group, foreign banks supported U.S. total credit growth during the recession, although many foreign banks, especially those from Japan, Italy, and the United Kingdom, cut back on loans over that
period. While Seth is unable to provide a full accounting of the continued strong loan
growth at foreign banks, she notes that their desire to increase market share and their
capital strength may have been important in maintaining the relative strength of foreign
bank lending.
The differing patterns of loan developments for foreign relative to domestic banks
have substantially reduced the domestic bank shares of total and business loans (Chart
3). Moreover, the flow of funds data used here understate the extent of foreign bank
loans to U.S. residents because offshore foreign banks' U.S. lending is excluded (McCauley and Seth 1992). Adjusted for offshore data, the true shares of U.S.-chartered
banks are considerably smaller than shown in Chart 3.
III. Factors Behind the Credit Slowdown
Studies in this volume investigate demand and supply factors underlying the slowdown
in private nonfinancial debt for both bank and nonbank sources of credit. The evidence
includes descriptive and econometric analysis and is based on hard data as well as on
survey materials for borrowers and lenders. On the demand side, the studies look for
both cyclical effects—the credit slowdown viewed as a by-product of the economic
slowdown—and noncyclical demand influences. On the supply side, the evidence for
both price and nonprice rationing of credit is considered.
A. Cyclical and Noncyclical Demand Influences
At an impressionistic level, the recent credit slowdown cannot be fully explained by the
1990-91 recession and the slow growth period surrounding the recession. Several studies in our collection—especially those by Cantor/Rodrigues, Lown/Wenninger, and
Mosser/Steindel—provide noneconometric data analysis of cyclical effects on various
debt or credit components. The general thrust of the authors' analysis of cyclical effects
is captured by data in Table 7, although collectively these studies cover a much broader
range of issues and detail. Briefly, the growth rate of private nonfinancial debt in nominal and real terms has been substantially lower in the period surrounding the recent recession than over comparable periods for the four earlier major recessions, on average,
or considered individually. Broadly, this pattern holds for major aggregate borrowing
components and for both bank and nonbank credit. The only significant exception is the
flow of home mortgage debt from both bank and nonbank sources, which has been significantly stronger in real terms over the period surrounding the latest recession than
around the last three major recessions since 1970.
The comparison of credit flows reported in Table 7 probably understates, to some extent, the contribution of cyclical developments to the private credit slowdown around
the current recession relative to the earlier episodes. As shown in Chart 4, the pace of
economic activity, nominal and real, was weaker in the current cycle than it had been on
average in the earlier cycles. Nevertheless, as pointed out by Lown/Wenninger and others, the differences in the pace of activity do not fully explain the sharp credit slowdown
in the current episode relative to the earlier episodes. Moreover, the credit weakness itself may be responsible, in part, for the slower pace of economic activity in the current
cycle. With changing relationships between credit flows and economic activity, it is
14
very difficult to assess the contribution of weaker-than-average growth in the current cycle to the severity of the credit slowdown. But one simple way to get a very rough sense
of this contribution is to use the average relationship between real credit flows and economic growth for the earlier cycles as a benchmark to calculate the implied credit flows
associated with recent growth performance. This type of exercise suggests that the
weaker-than-average pace of economic activity accounts for only about 35 percent of
the gap between the private credit growth in the current cycle and the average private
credit growth in the past four cycles.
Some noncyclical or structural demand shifts may also have contributed to reducing
the demand for credit in recent years. Such shifts are "permanent," by definition, but
their influence on demand may be difficult to separate from that of cyclical forces. Some
studies in this volume note the relevance of structural demand shifts in recent developments in credit flows. In particular, the Lown/Wenninger and Mosser/Steindel papers
discuss the influence of a possible downward shift in inventory demand relative to sales,
especially in the manufacturing sector, on the demand for commercial and industrial
Table 7: Credit Growth over Various Business Cycles3
Percent change, Annual Rate
Mortgage
Private
Business
Nonfinancial Nonfinancial
Household
Business
Home
Mortgage
Nonmortgage
Current dollars
Average,
current
cycle
3.1
0.7
5.6
-1.9
7.1
2.0
Average,
earlier
cycles(A)a
8.8
9.2
8.5
10.2
8.5
8.6
Average,
earlier
cycles(B)b
9.0
9.7
8.3
10.6
8.2
9.0
Constant 1987 dollars0
Average,
current
cycle
-0.4
-2.8
2.0
-5.4
3.6
-1.4
Average,
earlier
cycles(A)d
3.9
4.2
3.5
5.2
3.6
3.7
Average
earlier
cycles(B)a
3.2
3.9
2.5
4.7
2.4
3.2
a
Business cycle periods cover four quarters before trough, trough quarter and seven quarters
after trough.
b
Average of the 1958, 1970, 1975 and 1982 cycles.
c
Average of the 1970,1975 and 1982 cycles.
d
Based on GDP deflator.
15
Causes and Consequences
loans. Because of just-in-time and other management techniques, the amount of inventories needed for a given level of sales and, therefore, the financing requirements for
those inventories have declined in recent years. Even though such a shift is likely to
have been gradual and to have started before the recent credit slowdown, Lown and
Wenninger argue that a considerable portion of the unusual weakness in commercial and
industrial bank loans over the recent period may be explained by the need to finance a
lower-than-normal level of inventories.
Fxonometric analysis yields results which are broadly consistent with the less formal
data analysis, namely, demand influences as reflected in standard macroeconomic variables are unable, by themselves, to explain adequately the recent credit slowdown. At
the outset, it is worth noting that the estimates discussed here generally do not distinChart 4: Economic Activity in Various Business Cycles
Four quarters before trough = 100
-4
-3
-2
-1 Trough 1
Quarters before trough
Source: U.S. Department of Commerce.
16
guish between cyclical and noncyclical demand influences. The estimated equations
simply attempt to explain particular credit flows using aggregate demand components
and other appropriate macroeconomic factors as explanatory variables. Movements of
explanatory variables, in this context, capture all relevant normal or long-run influences
on credit flows.
Using cash flow and income or aggregate demand components as explanatory variables, Mosser and Steindel estimate total loan equations for nonlinancial corporations,
consumers, home mortgages, and business mortgages. They lind that swings in economic-activity-relatcd fundamentals seem to account for only about one-quarter to onehalf of the slowdown in corporate and consumer borrowings. In the case of consumer
credit, the authors reestimate equations by adding home equity lines to take account of
shifts between consumer credit and home equity loans resulting from the Tax Reform
Act of 1986; the results are roughly similar to those without the home equity variable.
For business and home mortgage components, estimates are unstable, although for
home mortgages, the estimated equations are able to explain the recent slowdown in
loans. Mosser and Steindel provide a particularly detailed analysis of corporate and
consumer loans and argue that most of the prediction errors for those loans do not seem
to reflect any exogenous shift in the relationships between credit demand and explanatory variables.
For bank loans, Lown and Wenninger estimate four sets of equations, one each for
commercial and industrial loans, business mortgages, home mortgages, and consumer
loans. The equations are estimated with VAR methodology to approximate reducedform relationships, using a range of economic activity and interest rate variables.
Broadly, the estimated equations for business mortgages and consumer loans underpredict the credit slowdown, while those for home mortgages more than fully account for
the extent of the slowdown. For commercial and industrial loans, Lown and Wenninger
are unable to reach any firm conclusions because of unstable regressions.
Cantor and Rodrigues estimate equations for total bank business loans and for nonbank business credit using GDP, investment, and inventories as explanatory variables.
The prediction errors from both the bank and nonbank equations are large, indicating
that macroeconomic activity variables do not provide an adequate explanation for the
slowdown in either bank business lending or nonbank business credit.
In summary, aggregate demand influences are unable to explain a substantial part of
the recent slowdown or decline in nonfinancial business borrowings from bank and nonbank sources; this is true for both mortgage and nonmortgage business borrowings. Demand factors also fail to account for the recent slowdown in consumer credit, and taking
account of shifts between consumer credit and home equity loans does not significantly
alter this result. Recent developments in total home mortgage debt and home mortgage
bank loans, however, appear to be adequately explained by the evolution of aggregate
demand influences.
B. Supply-Side Factors
With a significant fraction of the credit slowdown left unexplained by standard aggregate demand variables, one must turn to the supply side. Indeed, the prediction errors
or residuals from equations estimated with demand variables may be viewed as representing one measure of the supply-side influence on the credit slowdown. Of course,
even if we could account for all of the recent credit slowdown with the help of demand
variables, that result by itself would not necessarily imply that supply-side factors did
not contribute importantly to the credit slowdown. Such a result may simply reflect, for
17
Causes and Consequences
example, the fact that demand influences overwhelm supply-side factors. More generally, with both credit demand and supply falling, if the drop in credit demand is larger,
actual credit developments will tend to be dominated by demand influences, making it
difficult to estimate the net contribution of supply-side factors.
Four studies—Lown/Wenninger, Cantor/Rodrigues, Johnson/Lee, and Hamdani/Rodrigues/Varvatsoulis—in this collection have devoted considerable attention to the role
of supply-side factors in the credit slowdown. Their analysis covers bank and nonbank
sources of credit and survey data. Overall, the evidence points to significant credit supply problems for both bank and nonbank sources of credit.
On the bank side, Lown and Wenninger look at a number of supply-side factors and
provide both descriptive and econometric evidence on the role of those factors. They
find that in the 1989-92 period, spreads between bank lending rates and bank funding
costs for both corporate and consumer loans were at or above their previous record levels. They also note that the percentages of short- and long-term loans requiring collateral
increased sharply over 1989-92. Both indicators are consistent with a leftward shift in
the bank loan supply schedule.
Other noneconometric evidence discussed by Lown/Wenninger and others suggests
that banks engaged in nonprice credit rationing or, more generally, experienced reduced
ability or willingness to lend. Banks sharply increased their holdings of securities relative to loans, and some of the increase appeared to be noncyclical.6 Survey data from
banks indicate significant tightening in credit standards on mortgages and other business
loans during 1989-92.
Weakening bank capital positions—reflecting, in part, deteriorating bank loan quality and increasing charge-off rates—seem to have played a significant role in credit supply problems over 1989-92. Lown and Wenninger argue that poorly capitalized banks
reduced their lending more sharply than well-capitalized banks during 1990-91. Based
on a more comprehensive examination of the relationship between bank capital positions and bank credit, Johnson and Lee reach a somewhat stronger conclusion along the
same lines. Specifically, the results indicate that banks with weak capital positions did
less lending than banks with strong capital positions during the 1990-92 period.
Lown and Wenninger also argue that the increased emphasis by the regulators on
bank capital and the riskiness of bank loan portfolios may have contributed to the bank
loan slowdown, although the role of the regulators and examiners is difficult to separate
from other factors. While Lown/Wenninger and Johnson/Lee explore the effects of capital positions on bank lending, none of the studies in this volume explicitly investigate
the role of regulators and regulatory changes in the credit slowdown process.7
Using state-level data, Lown and Wenninger estimate cross-sectional regressions for
bank loan growth with employment, capital, and loan-loss reserves as independent variables; the latter two variables are intended to capture the effect of banking conditions
(that is, supply-side factors) on loan growth. The results suggest that capital and/or
lean-loss reserves contributed significantly to weak bank lending in 1990 and 1991, and
that the effects of these supply-side factors were greatest for the New Hngland region
followed by the Mid-Atlantic and the West South Central regions. By applying the
cross-sectional regression coefficients to changes in the explanatory variables by region,
18
6
More formally, Rodrigucs (1993) shows that weak economic activity cannot explain all of the recent runup in securities holdings and that the sustained steepness in the term structure of interest rates and riskbased capital standards may have contributed to that run-up.
7
For various perspectives on the role of regulators/examiners and capital standards, sec Greenspan (1992),
Syron and Randall (1992), Peck and Rosengren (1992), LaWare (1992), and Wojnilower (1992b, 1993).
Lown and Wenninger provide a quantitative sense of the contribution of supply-side factors to the overall bank credit slowdown. Specifically, they suggest that supply-side
problems accounted for roughly 15 to 40 percent of the slowdown in bank lending from
1989 to 1990.
Also using cross-sectional data, Demsetz estimated equations for bank loan sales
with expected economic activity, assets, capital ratios, nonperforming loan ratios, and
other bank characteristics as explanatory variables. She finds that both capital ratios and
nonperforming loan ratios are significant in explaining loan sales but their contribution
to predictions of loan sales declines is modest and swamped by that of economic activity.
Turning to nonbank credit sources, the Cantor/Rodrigues study offers evidence that
supply-side forces were at work here as well. The authors' econometric estimates for
nonbank business credit using GDP and its components as explanatory variables yield
large prediction errors which suggest a significant role for supply-side factors. The results also indicate that the timing of the credit slowdown for nonbank sources was parallel to that for bank sources, with no evidence of a shift from bank to nonbank sources
of funds.
Cantor and Rodrigues also provide considerable descriptive evidence on the role of
supply-side factors in the slowdown of credit from various nonbank sources—such as
finance companies, life insurance companies, and the commercial paper market. Business credit extended by finance companies advanced at a significantly slower pace starting in late 1989, when many finance companies were downgraded by the credit rating
agencies because of major losses in commercial lending and, more generally, weak balance sheet positions. With more credit downgrades during the recession and large
amounts put up for loan loss provisions and net charge-offs, total finance company business credit became roughly flat over 1990-92. Cantor and Rodrigues note that credit
downgrades probably had a significant effect on lending because finance companies
raise most of their funds in short-term public credit markets. The authors also suggest
that credit stringency at banks may have had adverse feedback effects on finance company credit availability as many finance companies, faced with problems in raising
funds in the commercial paper market, increased their borrowings from bank backup
credit lines, presumably at higher costs.
Most of the problems of the life insurance industry, Cantor and Rodrigues argue,
stemmed from commercial real estate lending, junk bond portfolios, and high rates on
guaranteed investment contracts. Against the background of weak economic activity,
these difficulties led to numerous credit downgrades, sharp declines in stock prices and
some outright failures in the life insurance industry. Life insurers became generally preoccupied with preserving liquidity and avoiding a collapse. In this environment, the National Association of Insurance Commissioners in mid-1990 adopted new rules
establishing more stringent reserve and capital requirements for below-investmentgrade bonds and private placements. These developments, Cantor and Rodrigues believe, have reduced the willingness of insurance companies to invest in below-investment-grade bonds and, more generally, have induced a shift toward low-risk assets.
Unlike the situation in earlier credit crunches, nonfinancial business borrowers did
not increase the rate of commercial paper issuance during the latest credit crunch. Because of numerous credit rating downgrades and fifteen defaults since 1989 (compared
with only two defaults in the entire earlier history of the market), perceived credit risk
in the commercial paper market increased greatly, leading investors, especially mutual
fund investors, to lose confidence. Meanwhile, to protect small investors and sustain
confidence in the money market mutual fund industry, the Securities and Exchange
Commission in July 1990 imposed strict limits on the amount of the "second-tier" (low-
19
Causes and Consequences
quality) commercial paper that mutual funds could hold. As a result of these developments, both the amount of second-tier commercial paper issued and the mutual fund
holdings of that paper dropped precipitously over 1990-92. Cantor and Rodrigues believe that the credit quality concerns are not fully reflected in the rate spread between
the top-tier and second-tier paper because the second-tier issuers arc often "rationed"
out of the market before they drive up rates. Cantor and Rodrigues also discuss the public bond market. The market for below-investment-grade public bonds ("junk bonds")
showed virtually no activity during 1990 and 1991 but recovered significantly in 1992.
By contrast, the market for publicly placed investment-grade bonds remained quite
strong, cushioning weakness in other credit markets to some extent.8
C. Survey Evidence on Supply-Side Factors
Hamdani, Rodrigues, and Varvatsoulis examine survey data from bank lenders and nonfinancial borrowers on credit tightening in recent years. Using both the narrative approach and econometric estimates, they find evidence of significant credit tightening by
lenders because of supply-side factors. By purging the NFIB (National Federation of
Independent Business) Survey data of aggregate demand influences, they uncover particularly strong and consistent evidence of a credit crunch for small business borrowers
that depend primarily on banks for their financing (about 90 percent of small business
debt consists of bank loans).9 The results indicate that the recent credit crunch for small
borrowers was more severe than earlier crunches. A significant part of this credit crunch
appears to have taken the form of nonpricc credit rationing or tightening of nonrate loan
terms.
Hamdani, Rodrigues, and Varvatsoulis also find considerable evidence of credit supply constriction for large borrowers. They find that overall, the extent of bank credit
tightening for large businesses appears to have been greater than what can be explained
by the general economic slowdown. Using the SLO (Senior Loan Officer) survey data
from banks, again purged of aggregate demand influences, the authors argue that the degree of credit stringency during 1990-91 seems to have been similar to that in the 197475 episode.
Finally, Hamdani, Rodrigues, and Varvatsoulis estimate loan growth models using
standard demand variables and survey variables on loan availability for both the SLO
and NFIB surveys. The results suggest that restrictive loan supply conditions as proxied
by the survey supply variables have had significant impact on commercial and industrial
bank loan growth over 1989-1992.
D. Correction for the Debt Overhang of the 1980s
As noted earlier, disentangling the supply and demand factors underlying the recent
credit slowdown is particularly difficult because the economic downturn was superimposed on a process of balance-sheet corrections for debt excesses of the mid-1980s.
This process of correction for earlier debt excesses is widely believed to have contributed significantly to the credit slowdown over 1989-92.
The severity of credit supply reductions, as noted earlier, has also been moderated somewhat by rapid
increases in off-balance-sheet lending (sccuritization and loan sales) in recent years.
9
20
In fact, the authors' credit supply proxies, purged of aggregate demand influences, may understate the
extent of credit supply shifts because they exclude supply shifts associated with movements of lending
spreads and at least some of the effect of changes in borrowers' quality on the willingness to lend.
During the last decade, a broad range of forces—including financial deregulation and
innovation, developments in information and data processing technology, commercial
real estate development, and mergers, acquisitions, and leveraged buyouts—combined
to increase greatly both the supply of and the demand for credit, resulting in enormous
increases in the amount of debt.10 The upward march of debt was supported, in part, by
speculative asset price increases, especially for real estate.
Over time, the process of rapid debt increases led, perhaps inevitably, to problems
for both borrowers and lenders. By 1989 and 1990, households and businesses faced
historically unprecedented and unsustainable debt and debt service burdens (Chart 5).
With declining real estate and other asset values, and weakening economic activity, high
debt burdens resulted in balance sheet difficulties for borrowers and loan quality problems for lenders. Not surprisingly, therefore, bank and nonbank lenders alike experienced a weakening of capital positions and increasingly higher loan loss reserves,
charge-offs, and delinquency rates. All these factors together, so the argument runs, explain the sharp credit slowdown in recent years.
This account of the correction process is consistent with the view that the credit
slowdown contained important supply-side elements although it was perhaps driven by
demand forces. In particular, in the down-phase, balance sheet changes induced by declining real estate and other asset values led to weaker capital positions for banks and
consequently lower capacity and willingness to lend over 1989-92, just as on the up-side
balance sheet changes had increased capacity and willingness to lend in the earlier period. The lenders' reduced willingness to lend, in this case, reflects not only changes in
their own balance sheets but also a shift in their attitude associated with the deterioration, actual or perceived, in the quality of borrowers' balance sheets and credit worthiness.
Perhaps even more important, according to this story, the correction process seems
to have been dominated by market forces (both demand and supply) as opposed to policy factors. In fact, monetary policy had been easing since early 1989, and as a result,
unlike earlier credit crunches, interest rates had declined significantly before serious
credit supply problems emerged. To be sure, tighter capital requirements and regulatory
pressures, stemming from both legislative changes and more intensive supervisory
oversight, contributed to the credit slowdown, in part by reinforcing and highlighting
prudential concerns. Such policy factors, however, appear not to have been the primary
cause of the credit slowdown. In any event, any contribution of policy factors to the
credit slowdown is likely to have been much smaller than the role played by market
forces; these forces particularly evident in a reduced desire to borrow and hold or extend
debt, caused a decline in both credit demand and credit supply."
Research work in this volume does not provide any estimates of the extent to which
the credit slowdown is attributable to the correction process for the debt overhang of the
1980s. While several studies discuss developments leading up to the credit slowdown,
quantitative assessments arc generally aimed at sorting out demand from supply (or cyclical from noncyclical) factors using historical trends. The study by Johnson and Lee
does address the related question of the linkage between the earlier credit excesses by
10
1
For a review of developments leading up lo the credit crunch period, see Cantor and Wenninger (1993).
For a broad perspective on the debt overhang of the 1980s, see Frydl (1991).
' Incidentally, note that shifts in attitudes toward debt would normally be treated as "exogenous" in most
macroeconomic models; the use of exogenous/endogenous in the current context, however, would appear
to be inappropriate since such terms must be expressed relative to a specific model.
21
Causes and Consequences
Chart 5: Debt Service Burdens
Percent
Source: U.S. Department of Commerce.
Notes: In the upper panel, debt service is an estimate of scheduled payments of principal
and interest on home mortgage and consumer debt. In the lower panel, cash flow is defined
as depreciation (book value) plus retained earnings (book value).
banks and the recent bank credit slowdown. It finds that banks that indulged in "high
risk" activities during the 1985-88 period were obliged to curtail their lending more
sharply than other banks during the three years to end-1992. But the study does not estimate the extent of "excess debt" resulting from those earlier high risk activities.
Nevertheless, it may be useful to get a rough sense of the impact of the correction for
the debt overhang on the credit slowdown since 1989. Specifically, I address the following question: Was the actual cumulative expansion in private nonfinancial debt from
22
end-1989 to end-1992 higher or lower than what is consistent with "normal" or long-run
trend credit growth adjusted for cyclical developments and for the debt overhang of the
1980s? Using the simple relationship that the amount of credit expansion in any given
time period is made up of the credit expansion consistent with the normal or long-run
trend rate adjusted for cyclical and other shifts away from that trend, I attempt to measure the gap between the actual credit expansion over the three-year period to the fourth
quarter of 1992, and the amount of credit expansion implied by the adjusted long-run
path under various assumptions for the relevant variables. If the actual credit expansion
over 1989-92 falls short of the estimated credit expansion for that period, the recent
credit slowdown has been greater than what could be reasonably attributed to the combination of cyclical effects and the correction for earlier debt excesses. In this case, the
correction process itself might have produced overshooting or shifts unrelated to the
earlier credit excesses, and cyclical developments might have further depressed credit
Hows. Of course, a significant positive gap between the actual and the estimated credit
expansion has the opposite implications.
There is no obvious and definitive way to measure the "normal" or long-run credit
expansion rate. The usual procedure is to use some measure of the historical trend rate.
But with credit expansion rates much higher in the 1980s than in the preceding two decades, history does not offer a clear choice for the trend rate or the benchmark period.
Perhaps more important, since long-run credit growth must be viewed in real terms, we
need relevant prices. At an empirical level, however, the choice of the appropriate price
measures needed to deflate various debt components is ambiguous. Similarly, the use
of the dcbt-to-GDP ratio at the component level in figuring out the long-run or normal
rate is quite problematical—the ratio of a particular debt component to GDP (or to broad
sectoral income measures) need not be stable over time. Adjustment of the long-run
trend to account for cyclical and noncyclical developments raises equally difficult questions: How should we measure cyclical effects? How much time should we allow for
the correction of the debt overhang to be completed—as much time as it took to build
up the problem, more time, or less?
Using various alternatives for the long-run or normal trend credit expansion rate and
adjustment factors, I calculated the cumulative amount of excess debt over 1982-89 and
several measures of the gap between the level of actual credit expansion over 1989-92
and the amount of trend credit expansion, adjusted for the debt overhang and cyclical
developments, during that period. One such exercise is reported in Table 8. The longrun trend rates in this exercise are based on business and household data for mortgage
and nonmortgage debt over the 1960-82 period, converted into constant 1987 dollars using the GDP deflator.12 The cyclical effects are measured on the basis of differences between the 1960-82 trend rates and the combined average growth rates for the periods
surrounding the 1970, 1975, and 1982 recessions.
This exercise suggests that the decline in business credit over 1989-92 has gone far
beyond what was necessary to correct the earlier debt excesses; only about 55 percent
of the decline in business credit over 1989-92 relative to the long-run trend can be attributed to the need to correct the debt overhang. Combining the correction for the debt
overhang with cyclical effects still accounts for only a part of the business credit slowdown. Hven assuming complete adjustment over three years (1989-92) for the credit ex-
The use of a national price index, instead of sectoral price indexes, seems to be preferable for at least two
reasons: appropriate component price measures arc not always readily available and. even when they are
available, their use would legitimatize credit excesses of the 1980s by incorporating any speculative price
increases for particular sectors such as real estate.
23
Causes and Consequences
Table 8: Long-Run Trend versus Actual Credit Expansion, 1989-92
Billions of 1987 dollars, GDP deflator basis
Business
Household
Total
Mortgage
Other
Total
Home
Mortgage
Other
Total
Private
Actual credit
expansion
-261.5
-159.0
-102.5
191.6
227.6
-36.0
-69.9
Trend expansion
423.6
151.5
272.1
284.9
185.9
98.9
708.5
Cyclical adjustment
-62.9
-2.0
-60.9
-93.5
-67.8
-25.6
-156.4
Correction for
excess expansion over
1982-89
-376.0
-75.1
-301.0
-665.1
-479.9
-185.2
-1041.1
Adjusted trend
credit expansion
-15.3
74.4
-89.8
-473.7
-361.8
-111.9
-489.0
Excess/
shortfall
-246.1
-233.4
-12.8
665.3
589.4
75.9
419.1
Partial adjustment
-461.0
-276.3
-184.7
285.2
315.2
-29.9
-175.8
Notes:
1) Changes in billions of 1987 dollars (GDP deflator basis) from 1989-1V to 1992-1V.
2) Long-run trends are based on the 1960-82 growth rates of business and household components.
3) Cyclical adjustments are based on the differences between the 1960-82 trends rates and
the combined average growth rates for the periods surrounding the 1970, 1975, and
1982 recessions.
4) Figures in the last row are estimated on the basis of partial correction (3/7) for the 198289 excess expansion over 1989-92.
5) In current dollars, actual cumulative private credit expansion over 1989-92 was about
$680 billion (11.0 percent of 1992 GDP).
6) Sums may not add up precisely due to rounding.
cesses that took place over seven years (1982-89), the actual business credit increase
over 1989-92 fell short of the long-run trend expansion, adjusted for the debt overhang
and cyclical effects, by about $246 billion in 1987 prices; the shortfall represents nearly
7 percent of total business credit at end-1992. Under partial adjustment, with three-sevenths of the excess debt eliminated over 1989-92, the debt shortfall from the trend expansion level increases to $461 billion, or about 12.5 percent of total business credit at
end-1992. While both commercial mortgages and nonmortgage business debt declined
more than implied by the estimated adjusted trend expansion levels under the two adjustment scenarios, the bulk of the shortfall reflects commercial mortgages.
For the household sector, the correction for the earlier debt excesses and cyclical effects, together, more than fully account for the credit slowdown. In fact, actual household credit expansion over 1989-92 exceeded the amount of credit expansion consistent
with the adjusted long-run trend, assuming complete adjustment over three years, by
24
$665 billion in 1987 dollars; the excess is nearly 17 percent of total household debt at
end-1992. About 90 percent of the excess debt is attributable to home mortgages. Under partial adjustment, the amount of household excess debt drops to less than half that
under complete adjustment, but it is more than fully accounted for by home mortgages,
with nonmortgage household debt actually showing a moderate shortfall relative to the
estimated level. In sum, there has been no correction for the debt overhang for home
mortgages. On the contrary, home mortgage debt over 1989-92 continued to advance at
a faster rate than the long-run trend rate, apparently unaffected by cyclical developments
and by the need for correction of earlier debt excesses.
Alternative measures of the long-run trend rate yield, in some cases, significantly
larger or smaller estimates of the debt excess over 1982-89 and of the gap between actual and estimated debt changes over 1989-92. Two general messages of the results in
Table 8 hold up, however. First, although the correction process for the debt overhang
played a major role in the credit slowdown, it is difficult to explain all of the business
credit slowdown by appealing to the need for correction. Second, home mortgage debt
in recent years has remained immune to the correction process for the earlier debt excesses. One implication of the first point is that some credit supply shifts largely or completely unrelated to the market correction process for the debt overhang may have
played an important role in the credit slowdown. Such supply shifts were presumably
caused by tighter capital standards and regulatory pressures.
E. The Credit Slowdown Abroad
A number of foreign countries have also experienced credit slowdowns, to varying extents, during the last three years or so. The Hickok/Osler study in this volume examines
the foreign experience, focusing on Japan, France, and the United Kingdom. Since a
single study cannot be expected to deal with all aspects of the foreign experience, the
authors consider only the broad contours of the recent credit experience abroad and the
common forces that may have driven that experience.
Using both descriptive analysis and regression results, Hickok and Osier (ind that for
all three countries, the waning of the credit surge of the 1980s contributed importantly to
the credit slowdown during 1990-91. The broadly defined process of financial deregulation and innovation, working through expanded access to credit markets, asset valuations
and other changes, led to increases in both the demand for and the supply ofcredit during
the mid- and late 1980s. Subsequently, as actual credit changes adjusted to "permanently" higher equilibrium levels, credit growth rates tended to return to more normal levels.
Hickok and Osier also find that for Japan and the United Kingdom, a reversal of the
speculative factors played a considerable role in the credit slowdown. Developments in
economic activity helped reduce the pace ofcredit growth in all three countries, but their
role appears to have been relatively modest in Japan and the United Kingdom. Finally,
bank capital movements seem to be significant in explaining credit movements in Japan
and to a lesser extent in the United Kingdom, but they appear not to have made any contribution to credit developments in France.
IV. Credit Supply Problems and Economic Activity
To the extent that the credit slowdown reliects the slowdown in aggregate demand or
economic activity, it is a symptom and not a direct cause of the weakness in the economy. Accordingly, any investigation of the consequences of the credit slowdown for nonfinancial economic activity must focus on credit supply problems. In this volume, three
25
Causes and Consequences
studies—Mosser, Stcindcl/Braucr, and Harris/Boldin/ Flaherty—deal with this subject.
Overall, the three studies clearly indicate that credit supply problems have not been the
primary or dominant cause of the recent weakness in economic activity. But, collectively, the studies do suggest that credit constraints are likely to have made at least some
contribution to the economic slowdown.
A. Aggregate Demand
Mosser examines the effects of credit supply problems on aggregate demand components while attempting to control for changes in credit demand. She estimates reducedform equations for several demand components with and without variables representing
credit supply restraints. Four different proxies, all based on other studies in this volume,
are used for credit supply constraints: (1) regression residuals from various bank loan
equations in Wenninger/Lown, representing part of the credit slowdown not attributable
to demand factors; (2) regression residuals from various sectoral loan equations in
Mosser/Steindel, measuring the gap between actual credit flows and the estimates based
on historical relationships between credit and aggregate demand variables; (3) residuals
from regressions in Hamdani/Rodrigues/Varvatsoulis, capturing credit availability restraints for small business, purged of cyclical influences; and (4) interest rate spreads between loan rates on business and consumer lending and market rates.
Using data for the 1980-92 period, Mosser performs some Granger-Causality tests to
determine whether credit aggregates or credit supply proxies are statistically more significant predictors of future economic activity. Her results tend to favor credit supply
proxies. For the more recent period, Mosser finds significant effects of credit supply
problems on commercial real estate activity and producers' durable equipment. In particular, the credit supply proxy for small business seems to account for a considerable
part of the 1989-92 weakness in nonresidential construction and producers' durable
equipment. F^ven so, Mosser argues that the weakness in these demand components relative to predictions based on normal historical relationships cannot be fully explained
by credit supply problems. Doubtless, the widespread sluggishness of economic activity during 1989-92 reflected a broader set of factors than just credit supply problems.
B. Construction Activity
Harris, Boldin, and Flaherty investigate the effects of credit supply problems on the real
estate industry. Focusing on the three construction industry sectors—single-family
homes, multifamily housing, and nonresidential structures—they provide a comprehensive review of credit and noncredit factors underlying the recent decline in construction
activity. Overall, the study finds that credit supply problems are likely to have played
only a modest role in the real estate contraction.
For single-family housing, the authors begin by examining predictions of housing
activity from several standard models that use mortgage rates, income, and other fundamentals as explanatory variables. Since these models are not able to predict the recent
weakness in housing, the authors search for an explanation by focusing on "special" factors or other variables that have been left out of the models. They argue that of the missing variables, demand-side factors such as a generalized effort to reduce debt and an
adverse shift in investor psychology rather than narrowly defined credit supply problems explain the bulk of unusual weakness in housing. This view is consistent with the
fact that because of the mortgage-backed securities market and other financial innovations, credit supply for home mortgages has not experienced any significant problems.
26
The supply of loans to homebuilders has been constrained significantly, but this appears
not to have caused a pervasive housing shortage. Even so, credit supply problems may
explain part of the recent weakness in housing activity since without credit constraints,
the housing supply would have been larger and prices lower. More generally, given the
weakness of both credit demand and credit supply, the identification problems make it
difficult to rule out a significant role for credit supply difficulties.
Multifamily and nonresidential construction have declined greatly since 1989 and
have remained as the two weakest sectors of the economy. According to the Harris/Boldin/Flaherty study, overbuilding in the 1980s (together with the resulting excess capacity) dominates the credit crunch as an explanation for the collapse of activity in both
sectors. The study recognizes, however, that these sectors have experienced credit supply problems and that the simultaneous weakness in (and interaction between) credit demand and credit supply makes it difficult to isolate the effect of credit supply constraints.
It is likely that in the absence of credit supply constraints, the decline in the nonresidential and multifamily sectors would have been more moderate. Put differently, the credit
crunch does not appear to be the dominant cause of the collapse in construction activity,
but it may well have played some role in the timing and process of decline.
C. Business Activity Excluding Construction
The Steindel/Brauer study explores the consequences of credit supply problems for
business activity excluding construction. Overall, this study provides only limited support for the view that credit supply problems impeded business activity over 1989-92.
Steindel and Brauer consider five different types of evidence. First, they review recent movements in corporate, noncorporate, and manufacturing activity, together with
relevant credit Hows. The review suggests that the sharp slowdown in credit flows may
have been a significant contributing factor to weakness in small business activity and
that such firms may have borne a disproportionate share of the shortfall in both output
and debt.
Second, the authors look at survey evidence on lending to smaller firms and the connection between credit supply proxies from other studies in this volume and noncorporate business output. This survey evidence does point to a significant credit tightening
which may have contributed to weakness in small business activity.
Third, using detailed industry- and firm-level data, the study compares activity for
small and large businesses and attempts to infer the role of credit in the recent weakness
of small business activity. The focus is on manufacturing businesses, but the analysis
does include some nonmanufacturing establishments as well. In most cases, small business activity appears not to have shown any unusual weakness relative to large business
activity, and so, by inference, Steindel and Brauer do not find any more support for the
effect of credit supply problems on small businesses than on large businesses. But with
data on the relevant credit flows unavailable, this type of evidence is entirely indirect
and does not necessarily contradict the view that credit supply problems may have contributed to the slowdown in business activity over 1989-92.
A fourth type of evidence considered by Steindel and Brauer focuses on indicators
of financial strength. Again using industry- and firm-level data, the authors explore the
role of financial factors in the recent weakness of business activity by examining various
measures of real economic activity for financially "weak" and "strong" businesses. This
evidence is also indirect and yields mixed results.
Finally, using firm-level data, Steindel and Brauer perform formal regression tests to
look for the effect of size and debt to asset ratios on employment, inventories, capital
27
Causes and Consequences
spending, and spending on research and development for various periods. Once again,
the results are mixed.
V. Implications for Monetary Policy
In reviewing the implications of the credit crunch or credit supply problems for monetary policy, this section focuses on two related issues: implications of the credit crunch
for the impact of monetary policy actions on economic activity, and consequences of
credit supply problems for monetary policy guides, M2, and other financial variables.
The section begins with some background information on the main features of the recent
credit crunch and on the channels of monetary policy influence on the economy.
A. Overview of Credit Supply Problems
The evidence in this volume is consistent with the view that credit supply problems contributed importantly to the credit slowdown over 1989-92, although demand influences
may have dominated overall credit movements. The nature and causes of the 1989-92
credit supply problems were significantly dissimilar to those of most earlier credit
crunches. The distinctive features of the most recent episode are summarized below.
First, credit supply problems in the 1989-92 period were widely spread across both
bank and nonbank sources of credit. As a result, unlike earlier credit crunches, nonfinancial borrowers were not able to substitute nonbank credit freely for bank credit. In
fact, finance companies, life insurance companies, and commercial paper issuance seem
to have experienced credit supply problems that were essentially similar to those of
banks. Together with a broadly based retrenchment in credit demand, credit supply
problems led to a sharp slowdown in all major components of private debt flows.
Second, credit restraints during 1989-92 took the form both of more stringent price
terms—higher lending rates relative to funding costs and tighter nonrate loan terms—
and of nonprice credit rationing. Although this phenomenon is probably fairly typical
of earlier credit crunches, the pervasiveness of nonprice rationing and tighter loan terms
over an extended period of time in the recent credit crunch is unusual. Earlier credit
crunches were generally short-lived; the 1989-92 crunch period was characterized by
persistently high spreads between lending rates and funding costs, especially at depository institutions, increasingly tighter credit standards for applications through much of
the credit crunch period, and continued stringent nonrate terms on loans. These persistent credit restraints were reflected, among other things, in large increases in holdings
of government securities relative to loans at banks.
Third, significant evidence points to a capital crunch as one of the major causes of
credit supply problems over 1989-92. None of the earlier credit crunches were characterized by a widespread weakening of the capital positions of banks and nonbank financial institutions. Broadly, the actual or perceived capital crunch seems to have reflected
three underlying forces (in addition to the normal cyclical effects): (I) the need to correct balance sheet problems resulting from the lax lending standards that had prevailed
through much of the 1980s and had left balance sheets badly exposed to asset prices and
other shocks; (2) increased capital requirements induced by legislative and regulatory
measures and by more intensive supervisory oversight; and (3) the weakening of capital
positions reflecting declining real estate and other asset values starting about late 1988.
Fourth, market forces seem to have played a critical role in generating the latest credit crunch. To be sure, as noted above, regulatory measures and pressures contributed to
the actual or perceived capital crunch but, unlike earlier credit crunches, the current ep-
28
isode emerged in an environment of accommodative monetary policy and declining interest rates. More fundamental to the process of credit slowdown appears to have been
the need to correct the debt excesses of the mid-1980s, which had become unsustainable
over time. Faced with major balance sheet and other difficulties, borrowers and lenders
alike responded to market forces, borrowers by lowering their credit demands and lenders by reducing credit availability. In particular, the so-called credit crumble phenomenon—the chain running from asset price declines to capital position weakness to lower
capacity and willingness to lend—contributed importantly to the process of credit slowdown. 13 The role of market forces was reinforced and perhaps intensified by the regulatory pressures which highlighted prudential concerns about loan quality and capital
positions and argued for the need to strengthen lenders' balance sheets. The capital
crunch itself was at least partly a by-product of the correction process as weakening capital positions and mounting loan losses called increasingly greater attention to the need
for correction of earlier debt excesses and for additional capital.
The accumulating loan losses, continuing balance sheet problems, and full realization of the debt overhang also led to more conservative lending attitudes—well beyond
what could be attributed to the measurable weakness in capital positions—and to a complete reversal of the earlier lax lending standards. To a considerable extent, the pervasiveness of credit supply problems reflected the widespread nature of the correction
process, with both bank and nonbank creditors experiencing the need to improve loan
quality and repair their balance sheets.
Finally, the debt overhang correction process and its conjunction with a prolonged
cyclical weakness in the economy made the already difficult task of identifying credit
supply malfunctions from credit demand factors even more difficult. Both borrowers
and lenders were deleveraging and restructuring their balance sheets in response to earlier debt excesses and to cyclical weakness. In the process, credit demand and credit
supply narrowed simultaneously, but the drop in demand is likely to have overwhelmed
the fall in supply. As a consequence, it is very difficult, if not impossible, to detect empirically the contribution of supply-side factors net of demand influences.
B. Channels of Monetary Policy Influence
Monetary policy influences the economy through at least four important channels: the
money-interest rate channel (or the "money" channel, as it is commonly known); the
credit channel; the asset valuations or balance sheet channel; and the exchange rate
channel.14 The discussion here deals with only the first three, ignoring the exchange rate
channel. In the money-interest rate channel, as enshrined in the standard I S - L M model,
monetary policy affects aggregate spending by raising (lowering) the cost of funds
through changes in the supply of money relative to the demand for money. Specifically,
monetary policy actions—open market operations, and so forth—induce changes in
bank reserves, money, short-term interest rates and, through substitution and expectational effects, long-term interest rates. Higher (lower) interest rates, in turn, raise (lower) the cost of funds, other things equal.
13
14
Sec Johnson (1991) lor a detailed analysis of this phenomenon.
A large number of theoretical and empirical studies on the transmission of monetary policy influence to the
economy have appeared since the mid-1980s. l:or some recent discussions of various channels of monetary policy, see Akhtar and Harris (1987), Bennett (1990), Bernanke (1993), Bernanke and Blinder (1988,
1992), Bosworth (1989), Friedman (1989), Gerllcr (1988), Gertler and Gilchrist (1992). Gerller and Hubbard (1988), Mauskopf (1992), Mosser (1992), and Romcr and Romer (1993).
29
Causes and Consequences
The credit channel, which may operate alongside the money-interest rate channel, affects aggregate demand through direct changes in the availability and terms of bank
loans. A tightening of monetary policy may reduce the supply of bank loans through
higher funding costs for banks or through increases in the perceived riskiness of bank
loans. Since the credit channel views bank loans as imperfect substitutes for other assets
(government securities, corporate bonds, commercial paper, and the like) in bank portfolios, monetary policy actions that reduce bank reserves and, therefore, deposits will
be matched by decreases in both securities and bank loans. As a consequence, borrowers with no access to other sources of credit will be obliged to reduce their spending,
while others with nonbank sources of credit, though less affected, will not be immune
to monetary policy influence as long as the alternative sources of credit are more expensive or less convenient.
The asset valuations channel of monetary policy influence on the economy works
through changes in balance sheet positions. Monetary policy actions that lower interest
rates, for example, tend to increase asset values and improve liquidity for firms by lowering interest-to-cash flow ratios. These balance sheet improvements, in turn, may increase business spending by raising the availability of internal funds and improving the
access to and the terms on external funds. Lower interest rates may also work to improve household balance sheet positions through debt restructuring and higher asset values, thereby increasing the availability of funds for debt retirement and additional
spending. Note that the argument of this channel is that interest rate changes may affect
spending by weakening (strengthening) balance sheets or wealth holdings, quite apart
from their effects on the cost of funds in the money-interest rale channel.
C. Effectiveness of Monetary Policy
Factors relating to the credit crunch seem to have created significant blockages for the
workings of all three channels of monetary policy. Overall, the blockages are likely to
have muted the impact of monetary policy actions on economic activity. The empirical
size and significance of the blockages are far from clear, however. Whether any of these
blockages will turn out to have permanent consequences for the conduct of monetary
policy is also not clear at this time.
The credit channel of monetary policy apparently was seriously disrupted over 198992. With the decline in the willingness and capacity of banks to lend, monetary policy
actions increasing bank reserves were not translated into additional bank lending. Specifically, easing of monetary policy apparently had very little impact on the supply of
bank loans over 1989-92. This view is clearly supported by increasingly tighter credit
standards, higher (or at least continued high) lending rates relative to funding costs and
restrictive nonrate loan terms. With nonbank credit sources also experiencing supply
disruptions, frustrated bank borrowers were not satisfied elsewhere. Much academic
discussion of the credit channel assumes that nonbank credit alternatives are easily
available to many (perhaps most) borrowers. This view clearly runs counter to the recent credit crunch experience. In fact, widespread nonbank credit supply disruptions
appear to have added substantially to the severity of the blockage in the credit channel.
The money-interest rate channel of monetary policy also seems to have experienced
some blockage during 1989-92. Policy-induced increases in bank reserves did translate
into lower short-term open market rates and faster growth of the narrow money, M I .
But the response of long-term interest rates and broader monetary aggregates to policy
actions was very sluggish and weak throughout 1989-92. The decline in credit supply,
as shown in the Hilton/Lown study, contributed importantly to slowing the growth of
30
M2. And presumably the shift in credit supply also played some role in maintaining
high long-term interest rates by putting upward pressures on rates, other things equal.
As a result, monetary policy actions were less effective in lowering the cost of capital,
hampering the workings of the money-interest rate channel.
The process of correction for earlier debt excesses may also have weakened the asset
valuations or balance sheet channel of monetary policy influence on the economy. Given the actual or perceived need to correct the large debt overhang, lower interest rates
may not have induced much additional spending by businesses and households because
the improvements in balance sheets and the underlying asset values materialized only
slowly. Put differently, easier monetary policy as reflected in lower interest rates may
have encouraged households and businesses to repair the perceived weakness in their
balance sheets by deleveraging and debt restructuring, without increasing spending significantly.
While credit supply problems during 1989-92 may have been important in reducing
the effectiveness of monetary policy, it is difficult to isolate their effects from those of a
broad range of other fundamental developments that are likely to have disrupted, weakened, or changed the linkages between monetary policy and economic activity. Mosser
discusses a number of these other fundamental developments. Of the factors not directly related to the credit crunch, the following appear to be particularly important:
• the response of long-term interest rates to short-term open market rates may have
been weakened by inflation fears or by a high level of investor uncertainty stemming from large federal budget deficits;
• effects of lower interest rates may have been weakened by very high levels of real
after-tax interest costs;
• looking from a longer term perspective, financial innovation and deregulation over
the last two decades are widely believed to have caused significant changes in both
the size and speed of monetary policy effects on various sectors of the economy.
Economic growth in recent years has also been restrained by factors unrelated to
both the credit crunch and monetary policy transmission—relatively tight fiscal policy,
a military build-down, excess capacity in the construction industry, and low levels of
consumer and business confidence. It is difficult to control for these nonmonetary influences in assessing the effectiveness of monetary policy.
Against this background, the quantitative significance of the 1989-92 credit supply
problems for the transmission channels of monetary policy is far from clear. As reported
by Mosser, econometric forecasting equations, both reduced-form and structural estimates from large models, significantly overpredict real spending from 1989 to 1992.
This finding is consistent with the notion that monetary policy actions have been less
effective in recent years than in the past. Presumably the overprediction reflects both
the credit crunch and other factors, however. Indeed, Mosser is unable to account for
all of the overpredictions by making use of credit supply proxies. Moreover, the overpredictions are not limited to sectors that are directly sensitive to monetary policy. Instead, they are widely spread across all sectors, suggesting a general malaise in
aggregate demand not captured by economic fundamentals.
Notwithstanding these measurement difficulties, credit supply problems during
1989-92 are likely to have contributed to reducing the effectiveness of monetary policy.
Clearly, the credit crunch weakened the credit channel and caused disruptions in credit
flows, producing at least some adverse consequences for economic activity. The credit
supply shifts are also likely to have hampered the workings of the standard money-in-
31
Causes and Consequences
terest rate channel and possibly to have weakened the balance sheet-related contribution
of lower interest rates to aggregate spending.
The long-term implications of the credit crunch for the effectiveness of monetary
policy are less clear. Recent credit supply problems may well cause durable changes in
the workings of monetary policy transmission channels by altering, for example, the relationship between changes in monetary policy and bank loans, between bank loans and
deposit Hows, and/or between debt and income.15 But such an outcome is by no means
certain. Moreover, with numerous other potential influences on the linkages between
monetary policy and economic activity, it may not be possible to isolate any permanent
traces of the recent credit crunch on those linkages.
D. Monetary Policy Guides
Disruptions in the linkages between monetary policy and the economy imply adverse
consequences for the usefulness of financial variables as monetary policy guides,
whether viewed as intermediate targets or simply as information variables. The usefulness of any monetary policy guide depends primarily on two considerations: the
strength and predictability of the relationship between the guiding variable(s) and the
ultimate objectives of price stability and economic growth, and the ability of the Federal
Reserve to define, interpret, and control the guiding variable(s).16 The recent credit
crunch seems to have added to problems on both counts.
Credit supply problems since 1989 have almost certainly contributed to reducing the
usefulness of M2 and M3 as policy guides. Hilton and Lown argue that the reduced willingness of depositories to lend was an important factor behind the weakness in deposits,
although their work does not fully isolate the effect of credit supply problems from that
of noncyclical credit demand factors. Specifically, the authors point out that relatively
high lending rates and the pervasiveness of stringent nonrate loan terms and nonprice
credit rationing reduced the supply of credit and, together with lower yields on deposits
relative to alternative assets, led to weak depository flows. Controlling for cyclical effects, Hilton and Lown estimate that by the middle of 1992, the credit slowdown had
lowered M2 growth by about 10 percent. Their regression results indicate that the
breakdown of M2 demand equations is at least partially attributable to the exceptional
weakness in credit formation; the predictive performance of M2 demand equations improves significantly when direct measures of credit or other factors capturing cutbacks
in lending are included as explanatory variables.
Credit supply malfunctions have also affected the relationship between credit aggregates and the economy. None of the studies in this volume is able to account for developments in various credit measures—household, business, bank and nonbank, and so
forth—over 1989-92 by using standard historical relations for macroeconomic variables. Of course, the underlying relationships of credit and monetary aggregates to prices and economic activity have not been particularly reliable during the last decade even
before the emergence of recent credit supply problems.
The usefulness of interest rates as information variables for monetary policy has also
been adversely affected by the credit crunch. With the pervasiveness of nonprice credit
rationing and stringent nonrate loan terms, changes in open market rates have had a
32
15
If, for example, the recent experience makes banks permanently more risk averse in their lending, monetary policy effects on bank lending would be smaller than before.
16
See Friedman (1993c) for a recent perspective on the role of financial variables in guiding monetary
policy.
smaller impact on credit conditions and economic activity than would otherwise have
been the case. Put differently, disruptions in the credit market mechanisms have made
past experience less pertinent as a reference point for understanding the effects of recent
interest rate changes on credit conditions and the economy. Similarly, to the extent that
credit supply problems influenced the yield curve and various interest rate spreads—
such as that between lending rates and funding costs or that between the (riskless) Treasury bill rate and the (risky) commercial paper rate—all these variables became less useful indicators, at least over 1989-92.
By reducing the information content of a broad range of financial variables, the credit
crunch has compounded the problems of finding appropriate guides for steering monetary policy. More specifically, credit supply problems in recent years have made it more
difficult to use M 2 or the federal funds rate (or any other financial variable for that matter) for determining appropriate money and credit conditions relative to the needs of the
economy. Even before the latest credit crunch, however, there was no significant agreement on the use of any one or two variables as monetary policy guides. Thus, the recent
experience with financial sector developments seems to have moved us further away
from a narrow focus on one or two intermediate targets toward the use of a broad set of
financial indicators as information variables to steer monetary policy.
VI. Some Concluding Observations
Collectively, studies in this volume offer evidence of a substantial, prolonged, and
broad-based contraction in credit supply over 1989-92. This finding strongly contradicts the view that the recent credit slowdown originated solely on the demand side.17
Research work reported here conclusively demonstrates that demand influences are unable to explain a significant part of the recent credit slowdown or decline in nonfinancial
borrowings from bank and nonbank sources. Moreover, the existence of credit weakness across a wide range of nonfinancial borrowings also challenges the notion that the
recent credit slowdown was nothing more than the bursting of a speculative bubble in
commercial real estate.18
The studies in this volume also indicate that the nature and causes of the recent credit
supply problems were markedly different from those of earlier credit crunches. In particular, unlike earlier crunches, the credit supply problems during 1989-92 were broadly
spread across both bank and nonbank sources of credit, with stringent loan terms and
nonprice credit rationing persisting over a relatively long period. Also, unlike earlier
episodes, the recent credit crunch was marked by a capital shortage and was driven, to
an important degree, by market forces. wSet in motion by the widespread balance sheet
difficulties of both borrowers and lenders, these market forces led to the correction process for the debt overhang of the 1980s.
The sharp, prolonged, and widespread decline in credit supply over 1989-92 would
be expected to have significant adverse consequences for the economy. It is therefore
not surprising that the credit crunch has sometimes been blamed for much of the weakness in economic activity since 1989. Yet, the studies in the volume do not support this
conclusion. On the contrary, they clearly indicate that credit supply problems were not
the primary or dominant cause of the weakness in economic activity over 1989-92.
Nevertheless, the studies do suggest, at least collectively, that credit constraints almost
17
Sec Mclt/cr (1991). and Klicson and Tulom (1992) lor particularly strong expressions of this view.
18
Sec, for example, Jordan (1992).
33
Causes and Consequences
surely made some contribution to that weakness and probably played a significant role
in slowing the economy before the recession and in impeding the recovery process.19
The apparent inconsistency between sharply reduced credit availability and its modest effects on economic activity is not hard to reconcile. The credit crunch has by no
means been the only factor depressing the economy. Other factors that contributed significantly to the 1990-91 recession and the subsequent weak recovery include the Gulf
war, the defense build-down, relatively tight (iscal policy throughout the period, generally high real long-term interest rates, low levels of consumer confidence, corporate restructuring, and the commercial real estate depression that followed the great build-up
of excess capacity during the 1980s. With so many powerful forces slowing economic
activity in recent years, one can hardly expect the credit supply problems to dominate
the picture. Moreover, the confluence of broad-ranging adverse influences on economic
activity and the market driven elements in the credit crunch make it difficult to isolate,
empirically, the effects of credit constraints on the economy.
Finally, this collection of studies suggests that credit supply problems over 1989-92
contributed to weakening the influence of monetary policy actions on the economy and
to reducing the usefulness of M 2 and other financial variables as policy guides. Whether
recent shifts in credit supply factors will have any long-term consequences for the conduct of monetary policy is far from clear, however. In the absence of further changes in
the regulatory environment, the long-term effect will depend, to a considerable extent,
on the durability of recent changes in attitudes toward debt on the part of lenders and
borrowers—specifically, whether lenders will continue to follow the recent risk-averse
approach to lending and whether the decline in the desired ratio of debt to income will
turn out to be permanent. The new conservative attitude toward debt may persist, but
such an outcome is by no means certain.
19
34
Perry and Schuliz (1993) and Friedman (1993b) reach a roughly similar conclusion.
References
Akhtar, M.A., and Ethan S. Harris. "Monetary Policy Influence on the Economy: An
Empirical Analysis," Quarterly Review, Federal Reserve Bank of New York,Winter
1987.
Bennett, Paul. "The Influence of Financial Changes on Interest Rates and Monetary
Policy," Quarterly Review, Federal Reserve Bank of New York, Summer 1990.
Bernanke, Ben. "Credit in the Macroeconomy," Quarterly Review, Federal Reserve
Bank of New York, Spring 1993.
Bernanke, Ben, and Cara Lown. "The Credit Crunch," Brookings Papers on Economic
Activity, 1992:2.
Bernanke, Ben, and Alan Blinder. "Credit, Money, and Aggregate Demand," American
Economic Review, May 1988.
. "The Federal Funds Rate and the Channels of Monetary Transmission,"
American Economic Review, September 1992.
Bosworth, Barry. "Institutional Change and the Efficacy of Monetary Policy," Brookings Papers on Economic Activity, 1989: 1.
Cantor, Richard, and Rebecca Demsetz. "Securitization, Loan Sales, and the Credit
Slowdown," Quarterly Review, Federal Reserve Bank of New York, Summer 1993.
Cantor, Richard, and John Wenninger. "Perspective on the Credit Slowdown," Quarterly Review, Federal Reserve Bank of New York, Spring 1993.
Dwight, Jaffee, and Joseph Stiglitz. "Credit Rationing" in Benjamin Friedman and
Frank Hahn, eds., Handbook of Monetary Economics, vol. 2, New York: North Holland, 1990.
Friedman, Benjamin. "Changing Effects of Monetary Policy on Real Economic Activity," Monetary Policy Issues in the 1990s, Federal Reserve Bank of Kansas City,
1989.
. "The Minsky Cycle in Action: But Why?" Quarterly Review, Federal
Reserve Bank of New York, Spring 1993a.
. Comments on "Was This Recession Different? Are They All Different?"
by Perry and Schultz, Brookings Papers on Economic Activity, 1993b: 1.
. "Ongoing Change in the U.S. Financial Markets: Implications for the
Conduct of Monetary Policy" in Changing Capital Markets: Implications for Monetary Policy, Federal Reserve Bank of Kansas City, 1993c.
Frydl, Edward J. "Overhangs and Hangovers: Coping with the Imbalances of the
1980s," Annual Report, Federal Reserve Bank of New York, 1991.
Gcrtler, Mark. "Financial Structure and Aggregate Economic Activity: An Overview,"
Journal of Money, Credit, and Banking, August 1988.
35
Causes and Consequences
Gertler, Mark, and Simon Gilchrist. "The Role of Credit Market Imperfections in the
Monetary Transmission Mechanism: Arguments and Evidence," New York University, unpublished paper, May 1992.
Gertler, Mark, and R. Glenn Hubbard. "Financial Factors in Business Fluctuations," Financial Market Volatility, Federal Reserve Bank of Kansas City, 1989.
Greenspan, Alan. Remarks at the 55th Annual Dinner of the Tax Foundation, New
York, November 18, 1992.
Johnson, Ronald. "The Bank Credit Crumble," Quarterly Review, Federal Reserve
Bank of New York, Summer 1991.
Jones, David M. "The Role of Credit in Economic Activity," Quarterly Review, Federal
Reserve Bank of New York, Spring 1993.
Jordan, Jerry L. "The Credit Crunch: A Monetarist's Perspective," paper presented at
the Annual Conference on Bank Structure and Competition, Federal Reserve Bank
of Chicago, May 7, 1992.
Kaufman, Henry. "Credit Crunches: The Deregulators Were Wrong," Wall Street Journal, October 9, 1991.
Kliesen, Kevin, and John Tatum. "The Recent Credit Crunch: The Neglected Dimensions," Review, Federal Reserve Bank of St. Louis, September-October 1992.
LaWare, John. Testimony before the House Committee on Banking, Finance and Urban
Affairs, July 30, 1992.
Mauskopf, Eileen. "The Transmission Channels of Monetary Policy: How Have They
Changed?" Federal Reserve Bulletin, December 1990.
McCauley, Robert, and Rama Seth. "Foreign Bank Credit to U.S. Corporations: The
Implications of Offshore Loans," Quarterly Review, Federal Reserve Bank of New
York, Spring 1992.
Meltzer, Alan. "There Is No Credit Crunch," Wall Street Journal, February 8, 1991.
Mosser, Patricia C. "Changes in Monetary Policy Effectiveness: Evidence from Large
Macroeconometric Models," Quarterly Review, Federal Reserve Bank of New York,
Spring 1992.
Owens, Raymond, and Stacey Schreft. "Identifying Credit Crunches," Federal Reserve
Bank of Richmond, Working Paper no. 92-1, March 1992.
Peek, Joe, and Eric Roscngren. "Crunching the Recovery: Bank Capital and the Role
of Bank Credit" in Real Estate and the Credit Crunch, Federal Reserve Bank of Boston, 1992.
Perry, George, and Charles Schultz. "Was This Recession Different? Are They All Different," Brookings Papers on Economic Activity, 1993: I.
36
Rodrigues, Anthony. "Government Securities Investment of Commercial Banks,"
Quarterly Review, Federal Reserve Bank of New York, Summer 1993.
Romer, Christina, and David Romer. "Credit Channel or Credit Actions? An Interpretation of the Postward Transmission Mechanism," in Changing Capital Markets:
Implications for Monetary Policy, Federal Reserve Bank of Kansas City, 1993.
Sinai, Allen. "Financial and Credit Cycles—Generic or Episodic?," Quarterly Review,
Federal Reserve Bank of New York, Spring 1993.
Syron, Richard. "Are We Experiencing a Credit Crunch?" New England Economic Review, Federal Reserve Bank of Boston, July-August 1991.
Syron, Richard, and Richard Randall. "The Procyclical Application of Bank Capital
Requirements," Annual Report, Federal Reserve Bank of Boston, 1991.
Wojni lower, Albert. "The Central Role of Credit Crunches in Recent Financial History." Brookings Papers on Economic Activity, 1980: 2.
. "Credit Crunch," in The New Palgrave Dictionary of Money and Finance,
New York: Stockton Press, 1992a.
. Discussion of "Crunching the Recovery: Bank Capital and the Role of
Bank Credit," by Joe Peek and Eric Rosengren, in Real Estate and the Credit
Crunch, Federal Reserve Bank of Boston, 1992b.
. "Not a Blown Fuse," Quarterly Review, Federal Reserve Bank of New
York, Spring 1993.
Wolfson, Martin. Financial Crisis, New York: M.E. Sharpe, 1986.
37
Causes and Consequences
38
Economic Activity and the Recent Slowdown in
Private Sector Borrowing
by Patricia C. Mosser and Charles SteindeV
In early 1990 debt formation in the United States—particularly loans made by intermediaries—slowed sharply. At roughly the same time, the economy entered a cyclical contraction, followed by a period of unusually sluggish growth. The relationship between the
slowdowns in activity and debt has been the focus of much debate: Did the sluggish economy "cause" a slowdown in credit formation via credit demand, or was slow economic
growth brought about by credit problems, particularly by restrictions in credit supply?
While such identification questions can rarely be answered definitively, this paper
looks at these issues in the context of the long-run relationships between credit flows
and activity. In particular the paper draws on two closely related empirical facts: (1)
Private debt formation goes up and down with movements in economic activity (see
Wojnilower 1980, 1985); and (2) the ratio of debt to output also has a cyclical component. By focussing on long-term credit/activity relationships, the paper attempts to
measure the fraction of the recent decline in private nonfinancial borrowing that may be
attributable to economic activity, as opposed to purely financial factors (such as borrowing to finance mergers and acquisitions [M&A| or leveraged buyout [LBO] activity) or
to temporary supply or "credit crunch" phenomena in debt markets.
The paper is organized into four sections. The first offers an impressionistic review
of the historical relationship between private debt formation and aggregate economic
activity. The second focuses on the linkages between business borrowing and business
activity, particularly in the corporate sector, and possible structural changes in business
spending which may have affected the credit/real activity relationships. The third carries out a similar analysis for the household sector. The fourth section touches on developments in the growth of mortgages. The broad conclusions of the paper are:
• Historical statistical relationships between business sector borrowing and a small
number of measures of business activity and between household borrowing and
spending fail to explain much of the movement in borrowing over the last decade.
In the mid-1980s, borrowing was substantially greater, and in the early 1990s, sub1
We thank R.G. Davis, M.A. Akhtur, Ethan Harris, and Richard Cantor for comments, and Cynthia Silvcrio
and Joshua Glcason for research assistance.
39
Causes and Consequences
stantially less, than the nonfinancial activity fundamentals predicted. Swings in the
fundamentals appear to account for considerably less than half of the decline in
short-term business borrowing from the mid-1980s through the early 1990s. In the
consumer credit market, shifts in the fundamentals may explain about half the falloff in borrowing.
• In the early 1990s several sectors of economic activity normally linked to debt formation were weak, even given the overall softness of the economy. These sectors
include consumer spending on discretionary items, capital spending on noncomputer
items, and inventory investment. Only in the case of inventories can a case be made
that this spending weakness was largely independent of credit market developments.
• The combination of a shortfall in borrowing based on historic debt/activity relationships and apparently related weakness in debt-sensitive spending suggests that a
significant fraction of the decline in lending was traceable to a contraction in the
overall supply of credit coupled with reduced demand for purely financial lending
(such as the decline in M & A and LBO).
• In business and household mortgage markets, relationships between borrowing and
nonfinancial activity were less stable. Swings in borrowing and activity were fairly
well linked over the 1980s, but the causation from credit supply to activity, or from
activity to credit demand was difficult to discern.
Economic Activity and Private Sector Borrowing
The growth of outstanding credit market debt owed by private U.S. nonfinancial entities
slowed dramatically in the early 1990s (Table 1). From 1983 to 1989 growth of this debt
averaged 10.9 percent (annual rate, not compounded), while from 1990-1 through 1992I I growth averaged a meager 3.5 percent. The slowdown was particularly dramatic for
nonfinancial business, corporate and noncorporate. Household debt growth also slowed
sharply to less than half the rate of the late 1980s.
The slowdown in debt growth, however, did not reverse the debt buildup of the
1980s. The ratio of private nonfinancial debt to GDP—which was near its early 1970s
level as late as 1983—rose sharply through 1990 and showed only modest declines
thereafter as the slow growth of nominal GDP offset the slowdown of debt growth
(Chart 1). The debt-to-GDP ratios of major private sectors—households and nonfinancial businesses—behaved in a roughly similar fashion.
Although much attention has been paid to the buildup and collapse of real estate
lending, Table 2 and Chart 2 show similar patterns in private debt excluding mortgages.
Table 1: Growth of Private Nonfinancial Debt
Percent Change, Annual Rate
Total3
Households
Business
Corporate
Noncorporate
1984-89
10.9
12.0
9.3
11.2
7.4
1990-92-11
3.5
5.4
1.1
2.0
-0.5
8434.5
3938.6
3593.3
2273.9
1319.4
1991-IV
Outstandingsb
a
- Includes state and local governments.
In billions of dollars.
b>
40
Chart 1: Private Nonfinancial Debt to GDP
Percent
160
The removal of mortgages makes little difference to the overall story; the buildup of
debt in the 1980s remains striking as does the subsequent retrenchment. In Table 3 we
look at the behavior of some of the most critical components of nonmortgage debt: consumer credit, long-term business and corporate debt, and short-term business and corporate debt. Short-term debt is further subdivided into bank loans and other. Much of
the weakness in debt growth from 1989-92 was due to outright declines in consumer
credit and bank loans to business, but there was also a dramatic slowdown in other nonbank short-term lending to business.
Charts 1 and 2 illustrate the relationship of debt formation to the business cycle. Other things equal, a faster pace of economic activity tends to increase both the supply of
and demand for funds in the credit markets. In particular, this cyclical element may reflect the close relationship between the demand for business and household credit and
the more cyclically sensitive components of real activity. For example, investment in
real estate, which is extremely cyclical, is very intimately connected to the debt markets,
while household spending on groceries, which is relatively acyclical, has a tenuous connection to debt formation.
Table 2: Growth of Private Nonfinancial Debt Excluding Mortgages
Percent Change, Annual Rate
Total3
Households
Business
Corporate
Noncorporate
1984 to 1989
10.2
8.3
9.9
10.9
6.0
1990 to 1992-11
2.7
1.5
1.8
2.5
-1.5
1991-IV
Outstandingsb
4397.2
1058.7
2435.9
2057.0
378.9
a
- Includes state and local governments.
- In billions of dollars.
b
41
Causes and Consequences
In addition, the charts highlight the boom/bust nature of debt formation during the
1980s and the early 1990s. In the 1980s expansion, debt grew very rapidly relative to
aggregate economic activity, while early 1990s saw an unusual drop in debt growth. A l though suggestive, by themselves these facts do not establish that debt movements were
"caused" by shifts in underlying supply and demand factors. Conceivably, shifts in the
composition of output first toward and then away from debt-sensitive activity could account for the wide swings in debt growth.
In light of the large differences in credit linkages across output components, the remainder of this paper focuses on specific sectors of real activity and movements in
closely linked types of borrowing. In particular, we first examine non-real-estate borrowing and debt-sensitive activity (capital spending, inventories, and the more discretionary elements of consumption). Near the end of the paper, we discuss trends in
mortgage lending and their relationship to nonfinancial activity, but we forgo discussion
of activity in real estate markets, which is handled in another paper in this study.
II. Developments in Nonmortgage Business Borrowing and Spending
Nonfinancial business debt covers a wide spectrum: bonds, mortgages, bank and thrift
loans, loans from finance companies and governments, commercial paper, and bankers
acceptances. Charts 1 and 2 showed the strong cyclical clement in debt formation in the
nonfinancial sector, with the ratio of debt to GDP typically growing during expansions
and falling during recessions. This trend was accentuated in the 1980s and early 1990s.
Tables I and 2 illustrate the marked deceleration in borrowing by nonfinancial business,
while Table 3 documents the very pronounced slowdown in short-term business debt.
Just as it is unreasonable to assume a fixed relationship between overall economic
activity and debt growth, so too is it implausible to expect a fixed relationship between
business activity and business debt growth. Certain types of business activity tend to be
more dependent on credit markets than others. For example, the demand by business
for nonmortgage debt is presumably related to business spending on inventories and fixed
Chart 2: Private Nonfinancial Debt Excluding Mortgages to GDP
Percent
42
capital. Although it is conceptually possible for firms to finance tangible capital formation solely through a combination of cash How and financial asset liquidation, it is extremely unlikely to occur: tangible assets are obviously ideal for (implicitly or explicitly)
collateralizing debt. Arguably, relationships between nonfinancial demand and business
debt growth could be examined at an even more disaggregated level. For example, the
demand for bank loans would likely be associated with inventory accumulation.
Moreover, there is good reason to confine formal statistical analysis to the corporate
sector, at least when discussing nonmortgage borrowing. A wider range of data are
available for the nonfinancial corporate sector than for the noncorporate sector (for instance, cash flow for noncorporate business is inadequately measured), and credit
growth in the noncorporate sector is very heavily influenced by real estate lending and
government programs for small business and farms. Chart 3 shows that recent trends in
noncorporate credit growth parallel those in the corporate sector; thus generalizations
drawn from analysis of corporate debt creation can likely be applied to the business sector as a whole.
One complication in examining the relationship between corporate borrowing and
activity is that the composition of the corporate sector's liabilities is subject to frequent
sharp changes for reasons that are not well understood and have little to do with economic activity.2 Since corporations are free to issue and retire equity, and in turn to retire and issue debt with the proceeds, there is some reason to look at the relationship
between nonfinancial activity and a broad measure of liability issuance broader than just
borrowing. The institutional changes in financial markets in the 1980s such as the
growth of the commercial paper, high-yield bond, and derivative markets, reinforce this
argument because they have blurred the distinctions between debt types and indeed between debt and equity.
" The Modigliani-Miller theorem holds that under certain circumstances, the demand for different types of
liabilities by a rational, profit-maximizing corporation will have no connection to its nonfinancial activity.
Merlon Miller (1978) has claimed that even under some "real-world" circumstances the theorem will
apply for individual corporations, although investor behavior will ultimately force the corporate sector as a
whole to some equilibrium capital structure.
Table 3: Growth of Selected Debt Categories
Percent Change, Annual Rate
Long-Term Debt
Nonfinancial Corporate LongTerm
Debt,
Consumer Excluding
Mortgages
Credit
a
Nonfinancial
Business
Bank Loans
Corporate
Noncorporate
Nonfinancial
Business
Other Short-Term Debt
Corporate
Noncorporate
Nonfinancial
Business
Corporate
Noncorporate
1984-89
8.8
13.0
10.2
12.8
8.0
6.2
7.2
3.9
10.3
12.4
7.7
199092-II
-0.8
5.8
2.4
4.4
-0.5
-2.6
-2.6
-2.7
0.5
-0.3
0.9
1991-IV
Ou (standings'1
796.7
1165.9
2323.4
1382.9
940.5
677.2
530.5
146.8
592.7
360.6
232.2
In billions of dollars.
43
Causes and Consequences
Chart 3: Nonfinancial Business and Corporate Debt Excluding Mortgages
Four-Quarter Percent Change
15
As a result, our strategy is to look at the relationship between business activity, specifically inventory and capital spending, and a number of liability measures for the nonfinancial corporate sector. We focus most of our attention on statistical models
excluding interest rates as determinants of demand. We do so for several reasons. There
is ample theoretical and empirical evidence (Stiglitz and Weiss 1981, Jaffee and Stiglitz
1990) that nonprice credit rationing is commonly used to sort out bad credit risks. In
such an environment, the "equilibrium" interest rate does not have a clear relationship
with the quantity of credit, in part because credit is allocated by means other than price.3
A second argument is the empirical one that increases in interest rates may depress the
demand for credit mainly through the indirect channel of depressing the level of debtsensitive activity.
We turn now to the details of the estimation procedure. As noted above, financial
and economic theory has not yet provided a satisfactory framework for building econometric models of the relationship between economic activity and balance sheet changes.
Some empirical models (Taggart 1977, Hendershott 1977, Auerbach 1985) have explored these relationships using stock-adjustment models. The rationale for such models is the hypothesis that households and businesses have desired long-run ratios of the
stock of financial assets or liabilities to flow measures of economic activity. In the short
run, the actual borrowing adjusts slowly to the long-run equilibrium. Unfortunately,
these models have not proven to be very stable—a finding that recalls the long history
of similar models of the demand for money.
Instead of looking at stock-flow relationships between the level of debt and activity,
we explore flow-flow relationships between borrowing—the change in debt—and activ-
A somewhat related but less technical argument advanced by Wojnilower (1980, 1985) is that the demand
for credit in the United States is insensitive to interest rates, and more importantly, that credit demand
(which is determined by real activity) exceeds supply at all "normal" interest rales. Wojnilowcr contends
that changes in credit supply availability drive credit flows and thus activity, almost irrespective of interest
rates.
44
ity. At a disaggregate level, we think it is likely that there are systematic (low-flow relationships stemming from customary methods of finance for certain types of activity.4
Breakdowns in such relationships—sizable errors in the simulated equations—may be
evidence of unusual factors at work in credit markets. Also, as a practical matter, flowflow models are simple to discuss (one is relating dollars of activity to dollars of borrowing), and in the absence of substantive theoretical or statistical evidence in favor of
alternative formulations, we prefer a simpler specification.
For corporate borrowing, we begin by using measures of current-dollar inventory
and fixed capital spending as major sources of demand. In addition to these spending
measures, corporate cash flow is likely to affect corporate borrowing. The relationship
here would tend to be more complicated than that between borrowing and spending on
inventory and capital. Higher cash flow implies less direct need to tap capital markets,
suggesting a negative relationship between borrowing and cash flow, but offsetting this
could be an association between higher cash flow and a greater ability to borrow. In other words, a negative coefficient on cash flow would suggest that demand effects outweigh supply effects, while the opposite conclusion would be drawn from a positive
coefficient.
The first measure of borrowing we examine is bank and thrift lending to nonfinancial
corporations. Equation 4.1 (Table 4) summarizes the model. The equation has a very
intuitive structure; corporate borrowing from intermediaries is negatively related to cash
flow and positively related to fixed capital and inventory spending. The residual standard error of the equation is about $20 billion, quite large compared with the mean of
the series.
4
The distinction between the stock-How and flow-flow specifications of debt demand can be reconciled at
an abstract level. Suppose the true underlying demand relationship is between stocks of debt and stocks of
income-producing capital (which can include human as well us physical capital). This true relationship can
be alternatively and cquivalcntly restated as one between slocks of debt and flows of income (assuming a
fixed relationship between flows of income and slocks of income-producing assets) or, if we lake first differences, as one between borrowing and changes in the stock of income-producing assets, or investment.
Table 4: Determinants of Nonfinancial Corporate Borrowing
Interest Rate Terms
Borrowing
Category
4.1 Bank and
thrift loans a
4.2 Short-term
debt b
4.3 All debt
4.4 Short-term
debt 5
Cash Flow
Plant and
Equipment
Spending
Inventory
Investment
-3.67
(.084)
.401
(.077)
.450
(.117)
-.531
(.145)
.663
(.136)
1.297
(.192)
-0.72
(.188)
.448
(.176)
.825
(.216)
-.793
(.479)
.715
(.442)
3.041
(.745)
Cash Flow
.048
(.045)
Plant and
Equipment
Spending
-.025
(-037)
Inventory
Investment
-.165
(.071)
S.E.
DW
R2
22.1
1.81
.231
.134
(.123)
33.8
2.00
.498
.385
(.114)
34.0
2.20
.736
.057
(.125)
33.4
1.99
.508
P
Notes: All data in billions of dollars, seasonally adjusted annual rates. Standard errors in parentheses. Estimation period: 1959-1 to
1992-11.
a
- Nonfarm nonfinancial corporate bank loans and S&L loans.
- Nonfarm nonfinancial corporate bank loans, S&L loans, finance company loans, trade debt, and commercial paper.
b
45
Causes and Consequences
Chart 4 shows the four-quarter moving average of loan extensions plotted against the
four-quarter change predicted by equation 4.1. From 1984 through 1988, borrowing
was about $ 10 billion to $20 billion a year higher than predicted by cash flow, fixed capital, and inventory spending. In relative terms, business borrowing was about 25 to 50
percent higher than the activity fundamentals predicted. Thereafter the situation
changed: in 1991 corporations on net liquidated loans from banks and thrifts at a pace
about $35 billion less than predicted by the fundamentals. Roughly speaking, the fundamentals suggest that by 1992 borrowing should have been positive at $10 billion to
$20 billion—comparable to levels seen in the aftermath of the 1981-82 recession. Instead, net borrowing from banks and thrifts remained negative.
Because many firms may substitute one type of corporate borrowing for another,
equation 4.2 broadens the focus to a larger debt aggregate.5 Added to bank and thrift
loans are finance company borrowing, commercial paper issuance, and trade debt. In a
qualitative sense, equation 4.2 is very similar to 4.1. The differences between the two
seem explicable—the greater sensitivity to capital spending may reflect the inclusion of
finance company lending (much of which is lease financing of equipment), and the
greater sensitivity to inventories may reflect the inclusion of trade debt. The standard
error of the equation is larger than that of 4.1, reflecting the greater average size of the
borrowing aggregate. However, the R2 of 4.2 is higher than 4.1, suggesting that some
of the noise in 4.1 reflects substitution into and out of the lending categories included in
4.2 but excluded from 4.1 .Chart 5 repeats for 4.2 the exercise shown in Chart 4. The results are similiar: borrowing was larger than predicted in the middle 1980s, but smaller
than expected in more recent years. As recently as late 1989 the actual and predicted
levels of short-term borrowing matched. Although the fundamentals suggest some softening in credit growth thereafter, the actual slowdown was much larger, with the model
5
Rcmolonu and Wulfkchclcr (1992), however, argue that there is a surprising degree of segmentation
between bank and finance company lending.
Chart 4: Bank and Thrift Loans
Actual versus Predicted
Billions of Dollars
46
Chart 5: Short-Term Borrowing
Actual versus Predicted
Billions of Dollars
150
overpredicting borrowing by as much as $75 billion. In fact, during 1991 the model predicted a positive level of borrowing about equal in magnitude to the actual rate of liquidation of short-term debt. In early 1992 the overprediction was still about $40 billion—
roughly comparable to the error in the bank and thrift loan equation.
Equation 4.3 expands the borrowing aggregate to include bond issuance (both standard corporate and tax exempt). In a qualitative sense, the results are similar to equations 4.1 and 4.2, with a negative coefficient on cash flow and positive coefficients on
the spending aggregates. However, the coefficient on cash flow is close to zero, suggesting that issuance of longer term debt may reflect forces other than demand conditions:
firms with strong cash flow may feel more confident about issuing long-term debt (a demand effect that is not associated with a cyclical need for funding). Equally likely, however, is the positive effect of strong cash flow on a firm's credit ratings and thus its ability
to attract long-term funds (a supply effect). Simulation of this equation showed once
again, that borrowing was greater than predicted in the mid-1980s and substantially less
than predicted in 1989-91.
For a variety of reasons, detailed analysis centers on the behavior of equation 4.2.
Equation 4.2 fits better than 4.1 because it covers a substantially broader aggregate:
short-term fluctuations from bank and thrift lending toward other short-term assets are
subsumed into the specification with the broader aggregate.6 The specification using an
even broader aggregate, equation 4.3, is superior in terms of overall fit, but is less clearly
reflective of aggregate demand factors.
The superiority of the short-term debt equation to the total debt equation is illustrated
in Table 5. This table shows the coefficients on cash How, plant and equipment spending, and inventories, for equations 4.2 and 4.3 when they are estimated over sample periods ending in 1979-1V (before the 1980 credit controls and the two recessions of the
1
Equation 4.2 was rccstimated without trade debt (which is not a component of credit market debt as usually
reported). The specification with trade debt was markedly superior on econometric grounds.
47
Causes and Consequences
early 1980s), 1983-1V (after the 1981-82 recession and just before the surge in debt formation of the mid-1980s), 1988-1V (after the debt surge and before the debt slump), and
the previously reported 1992-11. The relative stability of the coefficients in the shortterm debt equation is evident. Equation 4.2 appears to be picking up a stable long-run
relationship between borrowing and activity. The most obvious interpretation is that
equation 4.2 represents the fundamental or long-run influence ofnonfmancial factors on
the demand for short-term credit. The shifts in the coefficients for the total debt equation
suggest that the relative importance of nonfinancial influences and other effects (such as
corporate desire to restructure balance sheets between short-term and long-term debt
and between debt and equity) change erratically.7
7
Formal statistical tests arc generally supportive of the hypothesis that the structure of the short-term borrowing equation has been stable, while the structure of the overall debt equation appears to have shifted in
the late 1980s. Stability of coefficients does not mean that an equation fits particularly well, it merely
means that the influence of the factors considered docs not change.
A large number of other specification checks were done on equation 4.2. It is arguable that corporations have few options on the amount of short-term debt issuance, given cash How, plant and equipment
spending, and inventory investment. Thus, the good fit and stability of 4.2 might come from estimating a
near identity. However, removing cash How from 4.2 docs not greatly affect the equation's overall explanatory power, the pattern of its residuals, or the stability of the remaining coefficients. Heteroskcdasticity in
the residuals could be suspected; however, scaling the variables in the regression by G D P made no meaningful difference in the coefficients.
Table 5: Shifts in Activity Coefficients in Business Borrowing Equations
Cash Flow
Plant and Equipment
Spending
Inventory Investment
Short-term borrowing (equation 4.2 specification)
Sample period:
1959-1 to 1979-1V
-.875
(.231)
.966
(.198)
1.490
(.258)
1959-1 to 1983-1V
-.562
(.156)
.703
(.130)
1.344
(.199)
1959-1 to 1988-1V
-.294
(.133)
.497
(.120)
.883
(.192)
1959-1 to 1992-11
(equation 4.2)
-.531
(.145)
.663
(.136)
1.297
(.192)
All borrowing (equation 4.3 specification)
Sample period:
48
1959-1 to 1979-1V
-.742
(.179)
1.021
(.154)
1.332
(.221)
1959-1 to 1983-1V
-.600
(.124)
.867
(.103)
1.419
(.163)
1959-1 to 1988-1V
-.159
(.144)
.301
(.130)
.632
(.197)
1959-1 to 1992-11
(equation 4.3)
-.072
(.188)
.448
(.176)
.825
(.216)
We argued above that interest rate effects may not be important in explaining debt
growth when activity variables are taken into account. Equation 4.4 corroborates this
view. Interest rate effects are incorporated in the short-term borrowing model of equation 4.2 by adding variables that are the product of the measures of activity and the
prime lending rate to the basic equation. The logic behind this specification is a simple
demand story: higher short-term interest rates lower the sensitivity of borrowing to activity. At higher rates, a smaller fraction of inventory and fixed investment is debt financed, while a higher fraction of cash flow goes to retire short-term debt. As
hypothesized, the signs of these interactive terms are opposite to those on activity:
short-term rates affect short-term borrowing in a way consistent with a simple demand
hypothesis.8
Nonetheless, the addition of interest rate variables makes no real difference in explaining credit movements over the last decade. Despite formal statistical evidence that
interest rates directly affect the demand for short-term business credit (over and above
their effects on activity), the mid-1980s underpredictions of debt growth and the more
Footnote 7 continued
A variety of other borrowing aggregates-including the addition of equity issuance-were regressed on
the basic activity variables. None proved superior to the aggregate used in equation 4.2 in terms of the criteria of readily interpreted coefficients or stability. In any event, all models of borrowing screened showed
a consistent sharp swing toward ovcrestimation in the 1989-91 period. Some models estimated with end
points in 1988 captured the level of borrowing in the mid-1980s correctly but then grossly overestimated
the levels of the subsequent period, while others swung from large underestimates in the mid-1980s toward
overestimates afterwards.
8
An attempt to capture substitution between short- and long-term debt by adding long-term rates to the model
failed.
Table 6: Actual and Predicted Short-Term Borrowing, 1985-92
Billions of Dollars
Actual
Predicted
Actual less Predicted
1982
56.6
44.8
11.8
1983
63.3
53.4
9.9
1984
122.2
129.1
-6.9
1985
93.9
79.3
14.6
1986
66.5
63.0
3.5
1987
68.3
71.9
-3.6
1988
125.0
70.4
54.6
1989
98.3
90.5
7.8
1990
59.5
68.3
-8.8
1991
-36.5
36.7
-73.2
1992 (H-l)
-8.7
37.8
-46.5
1983 to 89 Average
91.1
79.7
11.4
1990-92 H-l Average
7.5
49.6
-42.1
49
Causes and Consequences
recent overpredictions were equally large for the models with and without interest rate
effects.
Table 6 looks more closely at the actual year-to-year movements in short-term borrowing and those predicted by equation 4.2. The swings in borrowing from 1982 to
1986 are well captured by the model; however, the surge in borrowing from 1986 to
1988 is understated. The decline in borrowing began in 1989, but equation 4.2 called
for an increase in that year. (The equation still underpredicted the level of borrowing.)
Equation 4.2 did predict a sharp decline in borrowing in both 1990 and 1991, but the
actual drop was much larger.
From the peak borrowing years of 1987-88 to 1990-91, corporate short-term borrowing declined $100 billion. The factors summarized in equation 4.2 explained only about
25 percent of this drop. The annual data in Table 6 tell a similar story for both sides of
the borrowing cycle: from the 1986 trough through the 1988 peak and back down to the
1991 trough the fundamental demand factors account for about a quarter of the movement in borrowing. In 1992, borrowing appeared to be returning to the fundamentals,
but the gap was still large.
We have documented that short-term lending to business in the early 1990s fell far
below what would have been predicted on the basis of the historical relationship between credit growth and business activity. The shortfall followed a period in the mid1980s when lending was well above this historical relationship. We now investigate
what factors may have contributed to this breakdown. One possibility is a simple exogenous shift in the demand relationship: the debt sensitivity of fixed and inventory investment may have increased, and the debt sensitivity of cash flow decreased. This
hypothesis is consistent with the underprediction of borrowing during the expansion of
the 1980s and overprediction of borrowing in the early 1990s. A second possibility is
an upward and then downward shift in credit supply. Finally, both credit supply and demand may have shifted up (in the mid 1980s) and then down (in the late 1980s) for reasons largely unrelated to real economic activity. For instance, the boom/bust cycle in
M&A and LBO activity may reflect such exogenous shifts in credit markets.
Chart 6A: Real Business Fixed Investment
Peak = 100
50
One way to distinguish between these possibilities is to look directly at the behavior
of the activity variables. A change in the structure of the underlying credit demand from
firms—for example, deleveraging as part of overall corporate restructuring—is likely to
be reflected in firms' nonfinancial decisions as well as their financial ones. Furthermore,
if restructuring is gradual, the changes in activity are likely to be gradual, hence gradual
changes in the pattern of business activity around 1989-90 could not plausibly be
charged to credit market constraints. Alternatively, if the slowdown reflected a restriction in the supply of credit, then one might see a sudden emergence of unusual weakness
in activity. Finally, if the cycle in credit creation was driven by shifts in purely financial
supply and demand factors, real activity might be unaffected.
The next section examines recent movements in two measures of business activity
that are closely linked to credit growth—fixed investment and inventory spending. Unusual weakness in these categories of spending may indicate that cutbacks in credit supply were restricting both the volume of borrowing and activity. If unusual weakness is
not apparent—or can plausibly be tied to developments independent of credit markets—
the implications for interpreting the slowdown in borrowing are less clear. Conceivably,
a gradual slowdown in activity may be consistent with a reduction in the demand for
debt. The third alternative discussed above—simultaneous shifts in both the supply and
(purely financial) demand for debt—does not have any straightforward implications for
activity.
Developments in Fixed Investment
A widespread impression exists that fixed investment held up quite well during the
1990-91 recession and its aftermath. This impression is only partly true. The decline
in constant-dollar fixed nonresidential investment in 1989-91 was less than was typical
for periods around recessions—even around mild recessions (Chart 6A). The overall
strength, however, hides some weaknesses.
First, the strength of investment was largely due to a surge in investment in comput-
Chart 6B: Real Business Fixed Investment Less Information Processing Equipment
Peak = 100
51
Causes and Consequences
ers and other forms of information-processing equipment. Removing computers leaves
real investment rather weak for a recessionary period, especially one with a relatively
high rate of capacity utilization (Chart 6B). This is more-or-less true even if we also
take out investment in commercial real estate (Chart 6C). The traditional parts of investment suffered a decided slump.
Second, the strength was in constant-dollar terms, not in current dollars. Overall,
current-dollar investment as a share of national output was low, and after 1989 its deChart 6C: Real Business Fixed Investment less Information Processing Equipment
and Commercial Construction
Peak = 100
105
Chart 6D: Nonfinancial Corporate Investment as a Share of GDP
Peak = 100
11
52
dine was in line with that seen in earlier recessions and their aftermaths. These patterns
appear to hold for the large share of current-dollar investment accounted for by nonfinancial corporations (Chart 6D). 9
What appears to have happened in the early 1990s was a surprisingly sharp cutback
in capital spending budgets relative to the mild economic decline. The real impact of the
cuts was blunted by the shift in the composition of capital spending toward computers
(whose prices were falling rapidly) and away from traditional, longer lived, and more
expensive types of capital. The abrupt shift in the composition of investment is at least
consistent with a restriction in the supply of credit.l() Such a supply cut could cause businesses to turn their attention toward more easily financed (and lower cost) investments
such as computers, while reducing spending on longer lived and more specialized types
of capital goods connected with long-term business expansion and modernization plans.
Developments in Inventory Investment
To a much greater extent than fixed investment, inventory investment experienced large,
fundamental shifts during the 1980s that were largely independent of the recent financial
distress. The most obvious manifestation of these changes was a very flat inventory cycle during the latest recession. During previous recessions, inventory investment accounted for more than three-quarters of the decline in real GDP, but in 1990-91, it
accounted for less than half of the drop in output.
Although the small inventory cycle in 1990-91 might have resulted from tight credit
supply conditions, the structural changes in inventory behavior appear to predate the
The detailed breakdown of investment by category is not available quarterly lor nonlinancial corporations.
10
This argument does not establish causality from credit supply to investment, however. The investment
shifts might be attributable to some other nonlinancial factor, while shifts in purely financial demand for
debt, possibly accompanied by supply changes, account for the cycle in borrowing.
Chart 7A: Inventory to Sales Ratio: Manufacturing and Trade
Months Supply
53
Causes and Consequences
Chart 7B: Real Inventory to Sales Ratios
Months Supply
2.1
"credit crunch." Chart 7A, which shows the monthly inventory-to-sales ratio for manufacturing and trade firms, suggests that the unusual inventory behavior during the recession was an extension of structural changes in inventory behavior, particularly a
downward trend in the stock/sales ratio, that started in the early 1980s." Furthermore,
changes in manufacturing inventory behavior were entirely responsible for both the
downtrend in stocks in the 1980s and the smaller cycle more recently.
In manufacturing, the introduction of "just-in-time"12 and similar inventory control
systems during the 1980s resulted in more effective and more frequent inventory control. For example, the percent of firms ordering production materials hand-to-mouth
(less than a week in advance) quadrupled, from 5 percent through most of the 1970s to
over 20 percent by 1991 (Chart 8A). 13 These innovations reduced both the average level
and the volatility of inventories, and thus can explain both the decline in manufacturing
inventory-to-sales ratios and the smaller inventory cycle in 1990-91.
54
1!
Recent revisions to constant-dollar inventory and sales data have moderated this downtrend. Upward revisions to manufacturing inventories over the 1980s and a sharp downward revision (of questionable origin)
to manufacturing shipments starting in 1990 raised recently published invcntory-to-sales ratios relative to
those reported in Chart 7A. The revised data suggest that structural change in inventories was quite pronounced in some industries but was less dramatic for all manufacturing.
12
As its name implies, just-in-time (JIT) is a production management technique advocating minimal inventories, frequent deliveries, and near-constant monitoring of production and inventory. Although JIT (which
works best when suppliers and buyers arc geographically close), may not be applicable to all U.S. manufacturing, most manufacturers have shortened their order leadtimcs by managing stocks more efficiently.
13
Anecdotal evidence suggests that the 20 percent of firms where JIT has become pervasive are in durables
manufacturing. Not coincidentally, these industries, which historically have very cyclical demand and
large inventory cycles, experienced much smaller inventory swings in the latest recession.
If the small inventory cycle was due to structural changes as opposed to credit restrictions, then a "real" model of inventory behavior, that is, one without any credit or
interest rate variables, should overpredict the small inventory cycle in the latest reces-
Chart 8A: Average Leadtimes for Production Materials
Percent Reporting "Hand to Mouth"
Percent
Source: National Association of Purchasing Managers. Shaded areas represent recessions.
Chart 8B: Dynamic Forecast of Manufacturing Inventory Investment Stock Adjustment
(Error Correction) Model
1980-88 Estimate
Change in Billions
6
55
Causes and Consequences
sion. Chart 8B shows dynamic forecasts for 1989-91 from such an inventory model for
manufacturing.14 Forecasted values for 1989-91 are consistently loner than actual investment, exactly the opposite of what one would expect to find if credit supply problems were hindering inventory investment.
While the move toward better inventory management appears to have been independent of credit supply shifts, it probably caused a downward shift in the demand for credit. The decline in corporate borrowing in 1989-91, however, was several times the size
of the decline in inventory investment, and thus inventory changes alone were insufficient to explain the weakness in business borrowing. Moreover, the downward shift in
inventory investment started in the early 1980s, but for much of subsequent decade corporate borrowing was considerably higher than would be explained by nonfinancial factors. Thus the connection between better inventory management and lessened credit
demand appears to be only a small portion of a much larger credit puzzle.
Developments in Household Borrowing and Spending
The bulk of household borrowing is done in the mortgage market. Households do, however, borrow to finance purchases of automobiles and other big-ticket items, and during
the 1980s, households increasingly used charge and credit cards to finance spending on
a wide variety of goods and services. This section examines the determinants of a major
category of nonmortgage household borrowing—installment and "other" (single-payment) consumer credit. This borrowing includes the most important types of nonmortgage household credit—automobile loans and credit card debt.
14
The real inventory model used for forecasts in Chart SB is an error correction model, which may be interpreted as a slight variation on the standard slock adjustment model. In this model, inventory investment is
determined by the gap between actual and target inventories (which are modelled as a function of sales) and
by lagged investment and sales changes. In addition, measures of JIT orders were included to account for
the structural changes over the 1980s.
Chart 9: Consumer Installment Credit Share of Disposable Income
Percent
190
56
Table 7: Determinants of Consumer Credit Borrowing
Disposable
Income
Discretionary
Spending
P
S.E.
DW
-.204
(.026)
.634
(.076)
.940
(.040)
8.71
1.92
Notes: All data in billions of dollars, seasonally adjusted annual rates. Standard errors in
parentheses. Estimation period: 1959-1 to 1992-11.
The sharp rise in the ratio of installment credit to disposable income in the 1980s and
its subsequent retrenchment (Chart 9) have been described as symptomatic of a household spending "binge" and "hangover" that inspired a return to more frugal habits. The
plunge in personal saving after 1984 and its modest revival after 1988 are also said to
be symptomatic of this phenomenon. As in the business sector, these large swings may
reflect credit demand shifts due to fundamental factors, credit supply constraints, or exogenous credit changes related to desired leverage.15 The model in Table 7 attempts to
explain household borrowing in the consumer credit market by means of the fundamental factors of income and spending. The income measure used in the equation is disposable personal income. The spending measure is consumer spending on discretionary
goods and services unrelated to home purchases. As in most of the business borrowing
equations, interest rate effects are excluded.
The equation tells us that historically a $1 increase in household income is associated
with about a $.20 decline in borrowing, while a $1 increase in discretionary spending is
15
In the household sector, unlike the business sector, exogenous shifts in credit demand, such as the desire to
increase and then reduce leverage, probably have substantial effects on activity, because households have
limited access to credit markets and no way to substitute "debt" for "equity."
Chart 10: Consumer Credit
Actual versus Predicted
Billions of dollars
100
57
Causes and Consequences
associated with roughly a $.60 increase in borrowing. The signs of the coefficients suggest that the equation is explaining credit demand phenomena. The standard error of the
equation is considerably smaller in magnitude than those for the corporate sector, but this
occurs in part because household borrowing is smaller in magnitude than business borrowing. Furthermore, the equation has a high degree of serial correlation, implying that
the right-hand side variables explain much less of the variance in the level of borrowing
than the reported standard error suggests. In fact, it is probably more correct to think of
this equation as explaining changes in, rather than the level of, borrowing.16
Chart 10 shows the four-quarter moving averages of actual and predicted household
borrowing. In essence, the story for household borrowing is much the same as for corporate borrowing: borrowing was stronger than that predicted by fundamentals in the
mid-1980s, more or less coincided with predicted levels in the late 1980s, and then fell
short of predicted levels in 1989-91.
Table 8 compares the actual level of consumer credit borrowing with that predicted
by the equation in Table 7. Consumer borrowing rose very sharply in the early 1980s,
peaking in 1984-85. From 1986 through 1989, annual borrowing fluctuated erratically
around $45 billion. Thereafter, borrowing fell off substantially, and by 1991 consumer
credit was being liquidated. Credit liquidation ended by mid-1992. Not surprisingly,
the equation missed the 1984-85 peak and instead predicted a smooth rise in borrowing
until 1989. As consumption weakened from 1989 to 1991, the model predicted about a
16
Rccslimulion of ihis equation over different sample periods produces little change in the coefficients. The
coefficients on spending and income for the 1959-79 sample period, in fact, are virtually the same as for
the 1959-92 period (the estimate of serial correlation, however, changes somewhat, and there is evidence
that the sensitivity of borrowing with respect to both income and spending was higher in the 1980-88
period than either before or after).
Table 8: Actual and Predicted Consumer Credit Borrowing, 1982-92
Billions of Dollars
58
Actual
Predicted
Actual less Predicted
1982
16.5
-16.9
33.4
1983
48.9
-0.1
49.0
1984
81.7
7.1
74.6
1985
82.3
17.9
64.4
1986
57.5
31.4
26.1
1987
32.9
46.1
-13.2
1988
50.1
55.6
-5.5
1989
41.6
58.8
-17.2
1990
17.5
44.4
-26.9
1991
-12.5
20.8
-33.3
1992 (H-l)
-7.8
24.4
-32.2
1983-89 Average
56.4
31.0
25.5
1990-92 H-l Average
0.4
31.0
-31.0
$40 billion drop in borrowing. The equation captured about three-quarters of the decline that took place over that period, but it consistently overestimated the level of borrowing. The model also tracked the change in borrowing in the first half of 1992;
however, the predicted level of borrowing was again sharply at variance with the actual.
Broadly speaking, the general pattern of consumer borrowing showed a sharp uptrend in the mid-1980s followed by a long-lived decline, with a particularly sharp decline in 1990-91. The nonfinancial fundamentals worked against the decline in the latter
part of the 1980s and worked to accentuate the 1990-91 plunge. After 1989, a large
share of the decline in borrowing—perhaps more than half—can be attributed to movements in the nonfinancial fundamentals, but it is clear that for the entire 1988-92 period,
economic fundamentals systematically overestimated the level of borrowing.17
As in the case of the business sector, we now examine consumer spending in the early 1990s for any abnormal changes that might have been associated with the shortfall in
borrowing. The unusual sluggishness in consumption was a major contributor to overall
economic weakness from 1989 to 1992. Compared to previous business cycles, the
more discretionary elements of consumption not related to home purchase were quite
sluggish from 1989 to 1992, comparable only to the period around the deep and pro17
The weakness in installment borrowing might have been an artifact of changes in the tax laws. The Tax
Reform Act of 1986 legislated a gradual elimination of the interest deducibility of nonmortgagc debt.
Other things equal, this change should have put some downward pressure on nonmortgagc borrowing (the
expansion of home-equity lending is a manifestation of this phenomenon). Conceivably, the severe weakness of nonmortgagc borrowing from 1990 through 1992 merely reflected the adjustment of borrowing
patterns to the new tax regime. As a rough check on the lax effect, the analysis was redone using a consumer borrowing measure expanded to include a measure of credit extensions on home equity lines.
Although this captures some of the tax substitution effect, it is not perfect because post-lax reform funds
raised from standard mortgage borrowing could also have been shifted to nonhousing use (when buying a
house, a household might borrow enough to also buy a car). Nonetheless, the reestimatcd equation produced coefficients very similar lo those shown in Table 7 and the inclusion of the home equity variable did
not greatly change the tracking ability of the model in the late 1980s and early 1990s: the actual decline in
household borrowing was still larger than predicted.
Chart 11 A: Real Discretionary Consumption
Peak = 100
108
59
Causes and Consequences
Chart 11B: Real Consumption
Peak = 100
106
Chart 11C: Real Nonmedical Consumption
Peak = 100
longed 1973-75 recession (Chart 11).18 Overall consumption was weak too, particularly
if medical care is removed.
The weakness in consumer spending over most of the 1990-92 period is little understood. Pronounced declines in indexes of consumer confidence and unusually small job
gains in the early stages of the 1991 -92 recovery are often cited as sources of weakness.
Discretionary consumption is measured as durable spending excluding major appliances, home furnishings, and medical durables, plus spending on restaurants, clothing, nonbusiness travel, other recreation,
and welfare and charity.
60
Chart 12: Actual less Predicted Real Consumer Spending
Billions of 1987 Dollars, Seasonally Adjusted Annual Rate
Even taking these factors into account, consumption was weak. Furthermore, there
were some positive developments—such as the strength of the stock market—which in
the past were associated with strong spending.
Statistical analysis of consumer behavior can be used to summarize much of" the evidence on the weakness in consumer spending. A recent version of the consumer sector
of the Federal Reserve Board model, where spending is essentially determined by income and wealth, showed that real consumer spending on nondurable goods and services swung from being $20 billion greater than predicted in 1989 to about $50 billion less
than predicted by early 1992 (Chart 12). Spending on motor vehicles, both in level and
growth, was also substantially less than predicted.
Unfortunately, it is unclear whether these large spending errors are closely linked (either as cause or effect) to changes in debt markets. For one thing, if consumer spending
were as strong as the Board model predicted, the personal saving rate would have been
near zero.19 In addition, it is arguable that the Board model—perhaps because of its
close linkage to the stock market and interest rates—is misspecilied and thus is misreading behavior. Indeed, the Board model had comparable overestimates of spending in the
19
Of course, if consumption had been as strong as ihe model predicted, there would have been favorable
effects on personal income and saving.
61
Causes and Consequences
Table 9: Growth of Private Mortgage Debt
Percent Change, Annual Rate
Corporate
Noncorporate
Farm
Nonfarm
Noncorporate
8.9
13.0
8.0
-5.3
10.2
7.2
-1.0
-3.0
-0.5
-0.8
-0.4
2879.9
1157.4
217.0
940.5
83.2
857.3
Total Nonfinancial
Households
Total
1984-89
12.0
13.7
1990-92-11
4.4
4037.3
1991-IV
Outstandingsa
a
- In billions of dollars.
1981 -82 recession. Further, consumer spending models (such as that of Data Resources
Incorporated IDRI]) that include consumer confidence measures did not overpredict.20
Given the weakness in consumer confidence and the softness in the job market, the
extent to which the consumer spending and the consumer borrowing shortfalls can be
traced to a cutback in the supply of credit to the household sector as opposed to a change
in the psychology of the consumer toward lower spending and credit demand is unclear.21
20
It is likely that the Board model overstates favorable stock market and interest rate development when consumer confidence is shaken by weak job markets; the DRI model (which is heavily dependent on the confidence index measures) underestimated spending in the early 1990s. Moreover, the Board staff reports
that a rcqent rcestimation of the Board model with revised data does a better job of tracking consumer
spending on nondurable goods and services, although the model continues to overestimate spending on
durable goods.
21
Even if consumers faced few direct restraints on their borrowing, but weak job formation due to credit supply constraints on firms' activity might have indirectly restrained consumer spending.
Chart 13A: Leverage on Owner-occupied Housing
62
Chart 13B: Leverage on Nonfinancial Business Real Estate
IV. Mortgage Borrowing
This section examines trends in mortgage borrowing by the nonfinancial sectors of the
U.S. economy. The actual trends in activity in residential and nonresidential construction are discussed elsewhere in this study; suffice it to say that in 1989-92 residential
construction was sluggish, and the nonresidential sector was extraordinarily weak.
Table 9 shows the growth of total, household, and business mortgage borrowing.
Home mortgage borrowing grew rapidly in the mid-1980s; growth slowed after 1989,
but no more so than other components of debt. Nonfinancial mortgages are quite another story, with extraordinary growth in the mid-1980s and outright declines in the 1990s.
Chart 13 shows leverage trends, both the ratio of home mortgages to the stock of residential real estate (13A) and the ratio of nonfinancial mortgages to the stock of nonres-
Table 10: Determinants of Private Mortgage Borrowing
Constant
Construction Spending
Cash Flow
Prorietors
Income
Noncorporate Capital
Consumption
10.1 Nonfinancial
business
21.250
(7.695)
1.582a
(.289)
-.035
(.120)
-.401
(.120)
-.817
(.496)
10.2 Household
home mortgages
-31.467
(3.671)
1.854b
(.122)
1
Disposable
Income
-.031
(.006)
P
R2
S.E.
.721
(.086)
.791
14.2
.918
24.5
Notes: All data in billions of dollars, seasonally adjusted annual rates. Standard errors in parentheses. Estimation
period: 1959-1 to 1992-11.
a
- Spending on nonresidential buildings, other nonresidential structures, plus nonhousehold residential structures.
- Household spending on residential construction.
b
63
Causes and Consequences
Table 11: Shifts in Activity Coefficients in Mortgage Equations
Business Mortgages
Cash Flow
Proprietors'
Income
Noncorporate
Capital
Consumption
.447
(.172)
.054
(.108)
.120
(.143)
.082
(.436)
1959-1 to 1983-1V
.065
(.256)
.166
(.089)
.085
(.115)
.029
(.362)
1959-1 to 1988-1V
1.023
(.297)
-.101
(.106)
-.124
(.115)
-.173
(.476)
1959-1 to 1992-1V
(equation 10.1)
1.582
(.289)
-.035
(.120)
-.401
(.120)
-.817
(.496)
Construction
Spending
Disposable
Income
1959-1 to 1979-1V
1.307
(.054)
-.019
(.004)
1959-1 to 1983-1V
1.391
(.047)
-.023
(.003)
1959-1 to 1988-1V
1.875
(.143)
-.035
(.008)
1959-1 to 1992-11
(equation 10.2)
1.854
(.122)
-.031
(.006)
Construction
Spending
1959-1 to 1979-1V
Sample period:
Home Mortgages
Sample period:
idential real estate (I3B). 22 Leverage on residential real estate increased in the 1980s,
stabilizing in the early 1990s. Leverage also increased sharply in the 1980s on business
real estate; furthermore, the upward trend continued in the early 1990s as the retirement
of mortgage debt failed to keep up with a continuing decline in values.23
Table 10 presents basic econometric models of business and household mortgage
borrowing. Business mortgage borrowing is modeled to depend on nonhousehold private spending on structures (excluding utility and energy structures), nonfinanciul corporate cash flow, proprietors1 income, and noncorporate depreciation.24 Household
64
22
These data value structures at replacement cost and land using an estimated market value.
23
The continued uptrend in this sector is notable since farm real estate and mortgages are included: the leveraging of farm property fell steadily after 1985 as debt was retired and values recovered.
24
As noted above, there is no measure of cash flow for noncorporate business. Proprietors' income cannot be
equated with after-tax profits both because it is gross of income lax and because the owner of an unincorporated business has great discretion over the fraction of its income stream that is recorded as profit-type
income as opposed to wages paid to the owner or insiders.
Table 12: Actual and Predicted Business Mortgage Borrowing, 1982-92
Billions of Dollars
Actual
Predicted
Actual less Predicted
1982
42.6
62.9
-20.3
1983
74.6
46.6
28.0
1984
91.8
57.7
34.1
1985
94.0
81.0
13.0
1986
99.1
60.0
39.1
1987
64.7
52.3
12.4
1988
55.5
51.5
4.0
1989
51.1
53.7
-2.6
1990
20.2
54.5
-34.3
1991
-17.6
5.8
-23.4
1992 (H-l)
-41.6
-10.9
-30.7
1983-89 Average
75.8
57.5
18.3
1990-92 H-l Average
-7.3
21.9
-29.2
mortgage borrowing is determined by disposable income and household spending on
residential structures. As in the corporate borrowing equations, interest rate effects are
detected statistically, but add little to the models' explanatory power or tracking ability.
In both equations 10.1 and 10.2, $1 of construction spending is associated with more
than $1.50 of mortgage borrowing. A coefficient greater than I may reflect an association between construction spending and increases in the values of, and turnover on, existing property. In equation 10.1, increases in noncorporate depreciation and proprietors'
income are associated with large declines in mortgage borrowing, while the relationship
with corporate cash flow is near zero. Increases in disposable income are also associated
with declines in home mortgage borrowing (equation 10.2).
Table 11 shows that equation 10.1 coefficients, both on income and spending, are unstable over time. In fact, during the 1959-83 period none of the fundamental variables
had a significant effect on borrowing. Clearly, and not surprisingly in light of the 1980s
boom-bust cycle, commercial mortgage borrowing is heavily influenced by factors other than spending and income fundamentals. Table 11 also shows signs of instability in
the home mortgage model (equation 10.2), although they are less pronounced than for
the business mortgage market.
The instability in these models makes it difficult to interpret their tracking abilities.
Table 12 shows that the year-to-year performance of equation 10.1 reflects the turbulence in this market. Only a small fraction of the runup in mortgage borrowing from
1982 to 1986 is explained by nonfinancial fundamentals, and actual levels of borrowing
in 1983-86 are understated by as much as 40 percent. Nonetheless, the fundamentals do
track a large portion of the collapse in mortgage borrowing in the late 1980s. The fundamentals suggest that borrowing should have fallen from $81 billion in 1985 to $6 bil-
65
Causes and Consequences
Table 13: Actual and Predicted Home Mortgage Borrowing, 1982-92
Billions of Dollars
Actual
Predicted
Actual less Predicted
1982
50.6
61.1
-10.5
1983
118.9
136.2
-17.3
1984
138.1
173.8
-35.7
1985
156.3
176.3
-20.0
1986
251.1
222.7
28.4
1987
260.3
234.1
26.2
1988
261.8
241.3
20.5
1989
251.9
252.1
-0.2
1990
223.8
196.4
27.4
1991
147.6
157.0
-9.4
1992 (H-l)
168.0
175.9
-7.9
1983-89 Average
205.5
205.2
0.3
1990-92 H-l Average
182.2
176.5
5.6
lion in 1991—a $75 billion decline. The actual decline over this period was about $110
billion. Still, the timing of the actual decline does not track the decline in fundamental
demand. In particular, the sharp decline in borrowing in 1990 cannot be explained by
the fundamentals, and the cumulative equation errors from the 1980s are not erased by
the end of 1991.
Even if we acknowledge that fundamentals capture the broad contours of the decline
in commercial mortgage borrowing, it would not be correct to infer that real activity
caused the drop in borrowing. This is one sector of the economy where causality is likely to be in the opposite direction: credit flows are likely to be major determinants of activity. Indeed, the instability in Table 11 suggests that equation 10.1 summarizes more
than just the influence of fundamentals on credit demand. A more rigorous analysis of
the impact of credit market changes on activity is needed in this sector to address the
issue of the relative importance of real activity and credit market factors.
Table 13 shows that, overall, equation 10.2 does a good job of explaining the runup
in home mortgage borrowing in the late 1980s and its slump from 1988 to 1991. Only
in 1984 was the difference between the actual and predicted levels of borrowing as large
as 20 percent. As in the case of commercial mortgage lending, however, it would be
improper to infer that the tracking ability of the home mortgage equation implies that
nonfinancial demand factors caused the drop in borrowing, because of the instability of
the estimated relationships and the likelihood that credit market factors strongly affect
demand. The most we can infer in both mortgage markets is that the declines in borrowing are correlated with the declines in activity. Of course, this conclusion leaves
open the possibility that shifts in credit supply helped cause the declines in activity.
66
V. Conclusion
Private sector borrowing in the early 1990s was quite weak, in dollar terms, relative to
the 1980s and—based on historical trends—relative to fundamental income and spending patterns. At the same time, several of the fundamental determinants of borrowing—
certain categories of fixed and inventory investment, and most categories of consumer
spending—were unusually weak given the overall cyclical context. The simultaneous
weakness in both borrowing and spending strongly suggests that the credit slowdown
did not result from a downshift in credit demand arising from purely financial factors.
The natural question arises of whether borrowing was weak because spending was weak
(low credit demand), or vice versa (reduced credit supply).
In the market for short-term corporate debt, nonfinancial activity factors accounted
for no more than one-quarter of the reduction in borrowing. At the same time, spending
(in current dollars) on fixed capital and inventories was also weak due to ongoing shifts
in inventory management techniques and in the composition of capital spending toward
high-tech equipment. These changes may have been part of the process of restructuring
U.S. business, and it is plausible to suppose that a downward shift in the demand for
credit was also part of that process. The shortfall in credit, however, was significantly
larger than could be explained by the restructuring process, and the weakness in both
spending and borrowing seemed too abrupt to be charged solely to a long-run restructuring process. Thus supply constraints in the credit markets, perhaps along with a
downward shift in credit demand for purely financial reasons, seem to account for much
of the shortfall in business borrowing.
In the consumer credit market, fundamental nonfinancial activity variables account
for perhaps half of the decline in borrowing from 1989 to 1992. Most of this period also
saw unusual weakness in household spending. While credit supply constraints could
have generated the both shortfalls, the plethora of explanations for the weakness in consumption suggests that much of slowdown in borrowing was due to demand factors excluded from our analysis. For example, extraordinarily slow job growth and low
confidence readings suggest that a shift in consumer psychology to less spending and
less borrowing also occurred.
The evidence is much harder to interpret for business and household mortgage borrowing. Declines in fundamentals such as construction spending superficially account
for the bulk of the borrowing decline in both categories; however, the relationships from
which we draw these inferences are quite unstable and are harder to interpret than those
above. The much closer links between credit and activity in these sectors suggest that
separate credit supply and demand influences are much more difficult to identify. A
more in-depth examination of these sectors is provided by Harris, Boldin, and Flaherty
in this volume.
67
Causes and Consequences
References
Auerbach, Alan J. "Real Determinants of Corporate Leverage." In B.M Friedman, ed.,
Corporate Capital Structures in the United States, Chicago (University of Chicago),
1985.
Hendersholt, Patric H. Understanding Capital Markets: Volume I: A Flow-of-Funds
Financial Model. Lexington: Ballinger, 1977.
Jaffee, Dwight, and Joseph Stiglitz. "Credit Rationing." In B.M. Friedman and F.H.
Hahn (eds.), Handbook of Monetary Analysis. Amsterdam: Elsevier, 1990.
Miller, Merton H. "Debt and Taxes." Journal of Finance, vol. 32 (March 1978),
pp. 261-75.
Remolona, Eli M., and Kurt C. Wulfekuhler. "Finance Companies, Bank Competition,
and Niche Markets." Federal Reserve Bank of New York Quarterly Review, vol. 17,
no. 2 (Summer 1992), pp. 25-38.
Stiglitz, Joseph E., and Andrew Weiss. "Credit Rationing in Markets with Imperfect Information." American Economic Review, vol. 71 (June 1981), pp. 393-410.
Taggart, Robert A. "A Model of Corporate Financing Decisions." Journal of Finance,
vol. 31 (December 1977), pp. 1467-84.
Wojnilower, Albert M. "The Central Role of Credit Crunches in Recent Financial History." Brookings Papers on Economic Activity, 1980:2, pp. 277-326.
. "Private Credit Demand, Supply, and Crunches—How Different are the
1980s?" American Economic Review, vol. 75 (May 1985), pp. 351-56.
68
The Role of the Banking System in the Credit
Slowdown
by Cara Lown and John Wenninger1
This paper explores the role of the banking system in the recent credit slowdown. In the
first section we briefly consider the economic developments leading up to the bank lending slowdown and argue that the slowdown was the result of four basic forces set in motion by the high inflation and high interest rates of the late 1970s. This first section also
contains a review of banking industry performance leading up to the bank lending slowdown. We show from several indicators of bank performance that a slowdown in bank
lending was emerging as a strong possibility by the late 1980s. The second section examines recent credit flows, analyzing the decline in bank lending both by lending and
by regional distribution. In the third section, we investigate the demand- and supplyside causes of the slower growth in bank lending. We find that only home mortgage
lending has been stronger in this cycle than in earlier cycles; all other categories of bank
lending have been weaker. However, the sluggish growth in commercial and industrial
(C&I) lending may reflect in part better inventory management, and may not be entirely
related to an increased reluctance of banks to lend. We also show that the increase in
the banking system's investment portfolio seems to be following, at least in part, a normal cyclical pattern. Some special features of this current cycle may have added to the
normal cyclical buildup in the banking system's holding of government securities, making it difficult to conclude that supply-side considerations are behind the increased holdings.
However, turning to other supply side considerations, we find several indicators of
reduced willingness to lend by the banks, including extremely wide interest rate spreads,
the results from the various surveys, a sizeable increase in the percentage of loans requiring collateral, and slower loan growth at banks with inadequate capital. This section
also contains a brief overview of the possible role of bank regulators in precipitating the
slowdown in bank lending from the supply side. The final section of the paper contains
an econometric evaluation of the recent sluggishness in bank lending. We present some
evidence that conventional reduced-form time series equations for bank lending break
1
We would like to thank Akbar Akhlar lor helpful comments on many drafts of this paper. Martina I level
and Joseph LaVorgna provided research assistance and Marie Krynicky prepared the manuscript.
69
Causes and Consequences
down during the most recent recession. In addition, our cross section results suggest
that problems in the banking industry contributed to weaker loan growth during the
1990-91 period, but not during the 1988-89 period.
I. Events Leading up to the Slowdown in Bank Lending
The recent slowdown in bank lending can be traced to four primary causes, all of which
have their origins, at least in part, in the high-inflation, high-interest-rate environment
of the late 1970s and early 1980s:
1) deregulation and innovation
2) the large buildup in debt burdens
3) the savings and loan crisis
4) a series of shocks to the banking industry stemming from losses on LDC loans,
agricultural loans, energy lending, and finally real estate loans and loans for highly leveraged transactions.
Because these four factors are highly interrelated, it is difficult to discuss their contributions in isolation. We will begin with a general discussion of the first three factors and
then turn our attention to the banking industry.2
The high rates of inflation in the late 1970s, building on a trend of higher peak rates
of inflation in each successive business cycle, seemed to leave open the possibility that
high inflation could return at some time later in the 1980s. Indeed, for much of the
1980s, core inflation, as measured by the consumer price index excluding food and energy, seemed stuck in the 4 to 5 percent range, creating doubts about the resolve of the
monetary authorities to reduce inflation over time through monetary control. In addition, the record budget deficits also added to the unease about the possible return of inflationary pressures. These expectations increased the willingness of firms and
consumers to raise their debt burdens and to invest heavily in nonfinancial assets, primarily commercial and residential real estate, but also in the stock market. From 1981
to 1986, commercial real estate also benefited from special tax incentives, an advantage
that helped to create a substantial oversupply of commercial buildings by late in the decade.3
On the lending side, expectations of rising asset prices and intense competition for
earnings, both domestically and internationally, prompted many lenders—banks, thrifts,
insurance companies and others—to lend on an expected asset (collateral) appreciation
70
2
A longer run historical perspective can also be found in Jerry L. Jordan, "The Credit Crunch: A Monetarist's Perspective," Federal Reserve Bank of Chicago's Annual Conference on Bank Structure and Competition, May 7, 1992; and Edward J. Frydl, "Overhangs and Hangovers: Coping with the Imbalances of the
1980s," Federal Reserve Bank of New York, Annual Report, 1991. More detailed reviews of financial
developments in the 1980s can be found in Thomas Simpson, "Developments in the U.S. Financial System since the Mid-1970s," Federal Reserve Bulletin, January 1988; and M.A. Akhtar and Betsy Buttrill
White, "The U.S. Financial System: A Status Report and a Structural Perspective," in C. Imbriani, P. Roberti, A. Torrisi, cds., // Mercato Unico Del 1992: Deregolamenlazione E Posizionamento Strategico
DeU'lndustria Bancaria in Europa, Bancaria Editrice S.p.A., Rome 1991, pp. 515-42.
3
For a detailed analysis of the tax law changes, sec James Potcrba, "Tax Reform and the Housing Market in
the Laic 1980s: Who Knew What, and When Did They Know It?", Real Estate and the Credit Crunch,
Federal Reserve Bank of Boston Conference, September 16-18,1992. For an analysis of why real estate is
susceptible to strong cycles even in the absence of tax incentives or disincentives, sec Lynn Browne and
Karl Case, "How the Commercial Real Estate Boom Undid the Banks," Real Estate and the Credit
Crunch, Federal Reserve Bank of Boston Conference, September 16-18, 1992. These authors refer to the
real estate cycle as a "hog cycle," or "overbuilding" caused by an inelastic short run supply curve and elastic long run supply curve" combined with multiyear leases that distort price information.
basis or on anticipated cash flow rather than actual cash flow and other measures of financial strength. This change in lending practices applied not only to real estate lending, but also to loans for highly leveraged transactions to acquire "undervalued assets."
The high rates of inflation in the late 1970s were also responsible for an increased
emphasis on financial innovation and deregulation, which lasted through much of the
1980s. These innovations not only enabled consumers and business to manage money
more efficiently and to earn market rates of return on their investments, but also to take
on increasingly large debt burdens and in some cases to access the money and capital
markets directly.4 Finally, innovation, deregulation and improved information technology created an environment in which there was far more intense competition for banks
on both the deposit and the lending side, resulting in reduced profitability and reduced
value of the banking franchise.
The regulatory response was not only to eliminate Regulation Q ceilings on the liabilities of banks and thrifts, but also to give the thrift institutions—already greatly weakened by the high interest rates—expanded asset powers without increasing supervision,
capital requirements, or lowering deposit insurance coverage.5 This regulatory response
enabled thrifts, with little experience outside the home mortgage lending area, to invest
heavily in risky assets, including commercial real estate. Here again, financial innovation played a role. On the liability side, brokered deposits, issued in insured units of
$100,000, enabled weak thrifts to tap the national money market with a managed liability. On the asset side, the high yields on junk bonds seemed very attractive to recently
deregulated thrifts at the same time that it became relatively easy for other types of lenders to originate and sell mortgage loans, the traditional thrift asset, reducing the need for
a specialized home mortgage lender. The end result, of course, was the massive "thrift
bailout" of the late 1980s and early 1990s, which focused the attention of the public on
the health of the nation's financial intermediaries (in terms of where "the next problem"
might be) and created a more cautious lending environment.
Against this background of innovation, deregulation, inflationary expectations,
heavy debt burdens, a weak thrift industry, and substantial overbuilding in the commercial real estate market, we review the performance of the banking industry. Table 1 presents some selected indicators of commercial bank performance before and after 1988
that may have had some impact on the willingness of banks to lend after 1988. The first
two columns show that by the late 1980s, consolidation in the banking industry was becoming more apparent. The number of institutions began to shrink rapidly, and average
assets held per institution increased steadily. In addition, the percentage of unprofitable
banks hit a peak in 1986-87 (column 3). The fourth column shows the large increase in
the number of failed banks by the late 1980s, a number that peaked at 220 in 1988.
4
Henry Kaufman has argued that inflation (or inflationary expectations) by itself is loo simple an explanation for the rapid growth of debt in the 1980s. He also points to a shift in attitude toward debt, financial
innovation (including securiti/alion), deregulation, financial internationalization, the lax structure, and the
practice of debt prudence. Kaufman's views arc summarized in "Debt: The Threat to Economic and
Financial Stability," in Federal Reserve Bank of Kansas City, Debt, Financial Stability and Public Policy.
August 27-29, 1986, pp. 15-26; and presented in more detail in Henry Kaufman. Interest Rates, the Markets, and the New Financial World (New York: Times Books), 1986. Not everyone shares this view, however. Benjamin Friedman takes the position that "at least for the present, ...the most honest answer of why
all this [debt acceleration| has happened in the 1980s is that nobody really knows" ("Changing Effects of
Monetary Policy on Real Economic Activity," Federal Reserve Bank of Kansas City, Monetary Policy
Issues in the 1990s, August 30-Scptcmbcr 1, 1989, p. 70).
•* For more detail on the thrift crisis, see Edward Kane, The S&L Mess: What Really Happened? Boston,
MA: MIT Press, 1985; and Lawrence White, The S&L Debacle:
Public Lessons for Hank and Thrift Reg-
ulation (New York, NY: Oxford University Press, 1991).
71
Causes and Consequences
Moreover, by the late 1980s, asset quality problems beyond the less developed country
(LDC) lending problems encountered early in the 1980s were becoming very apparent.
For example, the annual review of "Recent Developments Affecting the Profitability
and Practices of Commercial Banks" in the Federal Reserve Bulletin for 1986 notes:
• Moreover, loan losses continued to rise, and the sharp drop in oil prices and associated dislocations in the energy sector, as well as continuing weakness in agriculture
and overbuilding in commercial real estate, all point to lingering—and possibly
worsening—problems with asset quality for many banks.
Reflecting these problems, the table (column 5) shows that compared with the early
1980s, the period beginning in 1987 saw a dramatic downward shift in banks' average
return on assets. The 1987-88 period also saw a substantial weakening in equity capital
positions for banks, a development that led regulators to encourage higher capital ratios
and larger loan loss provisions. As Table 1 (column 7) indicates, loan loss provisions
increased dramatically after 1986, and equity capital ratios fell sharply until 1989, when
the capital positions of the banking system began to improve. The return on assets was
quite weak on average from 1987 to 1991; a considerable number of banks remained on
Table 1: Selected Measures of Bank Performance
All U.S. Insured Commercial Banks
Number
of
FDIC Problem
Return
Loss
Equity
Banks
on
Return Provision Capital
Percent
Assetsb
Unprofit- Number Assets on Equity (Percent- (Percent(In
(In
of Failed
(In
age of
age of
able
Banks Percent) Percent) Assets) Assets) Number Dollars)
Banks
Banksa
Assetsb
1981
14,208
1,940
5.22
7
.76
13.10
.26
5.83
196
n.a.
1982
14,123
2,101
8.27
34
.7
12.10
.40
5.87
326
n.a.
1983
14,075
2,245
10.58
45
.67
11.24
.48
6.00
603
n.a.
1984
13,952
2,401
13.06
78
.64
10.59
.57
6.01
800
n.a.
1985
14,398
2,593
17.09
118
.70
11.07
.70
6.21
1,098
174
1986
14,202
2,799
19.79
144
.63
9.84
.81
6.25
1,457
286
1987
13,701
2,962
17.66
201
.09
1.40
1.30
6.08
1,559
329
1988
13,134
3,097
14.65
221
.80
12.98
.58
6.06
1,394
305
1989
12,727
3,234
12.50
206
.49
7.59
.98
6.26
1,092
188
1990
12,373
3,391
13.43
159
.49
7.57
.96
6.31
1,012
342
1991
11,954
3,434
11.59
108
.53
7.86
1.02
6.57
997
528
1992
11,499
3,497
6.75
100
.92
12.80
.77
7.06
787
408
Sources: Federal Reserve Bulletin and FDIC Quarterly Banking Profile.
a
- The data for the number of banks before and after 1985 are not comparable. The Board staff revised this series
recently but only as far back as 1985. The result was to add 350 to 400 banks to the total. Hence, there was no
"peak" in the number of banks in 1985 but probably a continuous downward trend throughout the period shown
in the table.
b
- Dollars amounts are in billions.
72
the FDIC's problem bank list during that time, and by 1990 larger banks were emerging
as the distressed banks.
Profits fell further in 1990 from the low levels of 1989, and the banks were under
pressure to increase their loan loss provisions as more and more loans became nonperforming. In 1990, however, the loss provisions were for commercial real estate lending
and merger-related credits rather than loans to developing countries, agriculture, or the
energy sector as had been the case earlier in the 1980s. Bank equity prices declined, investors demanded larger risk premiums on subordinated debt, and some large banks had
difficulties in funding themselves in the interbank market. Banks restricted their growth
in anticipation of the new capital standards, maintained wide margins on loans, tightened credit standards, increased loan securitization, and cut operating costs. 6 At the
same time, the banks were also aware that the excesses of the thrift industry were being
resolved on a massive scale, perhaps reducing the size of the industry by one-third or
more; this process focused public attention on the soundness of other financial intermediaries and may have made banks more cautious.
In sum, this review of the performance indicators in Table 1 suggests that the banking
system was suffering severe distress in the late 1980s that could well have contributed
to reduced credit availability from the supply side. Not until 1992 were there some indications that the problems at banks had stabilized. Banks increased their equity and
debt issuance, stopped tightening their credit standards, and in some cases, gave signs
of easing. Moreover, with the credit ratings of many LDCs greatly improved, it appeared that some large banks had over reserved for losses in LDC loans, and these "excess reserves" could be transferred to support other troubled loans. In the next three
sections of this paper, we assess more fully the role of the banking system in the recent
slowing of credit.
II. Recent Trends in Bank Credit Flows
Table 2 shows the broad credit flows over the 1960-91 period, with the time since 1980
broken into three-year intervals to capture some of the shifting trends during the 1980s.
From 1960 to 1979, total debt increased at about the same rate as GDP, while depository
credit grew about 1.0 percentage point more rapidly than GDP. In 1980-82, these trends
began to change. Total debt accelerated and began to grow more rapidly than GDP,
while depository credit—primarily at thrift institutions—slowed sharply as home mortgage lending came to a virtual halt. But as the economic recovery progressed and consumers and corporations became more willing to take on debt, total debt as well as bank
and thrift credit accelerated sharply. In the 1983-85 period, the growth of total debt exceeded GDP by 4 percentage points, depository credit growth exceeded GDP by
2 percentage points, and nondepository credit exceeded GDP growth by more then
6 percentage points. Nondepository credit growth has continued to exceed GDP growth
by a wide margin (4.5 to 5.5 percentage points). Depository credit, on the other hand,
decelerated sharply as thrift credit went into an outright decline in the 1989-91 period.
Relative to the peak growth rates of the 1983-85 period, total debt decelerated
7.0 percentage points, nondepository credit about 5 percentage points, and depository
6
A more detailed review of the problems banks encountered during the 1980s can be found in the testimony
of John La Ware before the Committee on Banking, Housing, and Urban Affairs. U.S. Senate, June 10,
1992.
73
Causes and Consequences
credit 11 percentage points.7
The trends in bank and thrift credit, as indicators of the supply of and demand for
credit, have been distorted by several developments in the 1980s. The decline in thrift
credit probably does not have economic consequences commensurate with its size.
With a large share of single-family mortgages being securitized and sold in the capital
markets, mortgage money has remained available to consumers at market prices even as
the thrift industry has downsized. In addition, in the case of failed thrifts, the remaining
assets have either been funded and held by the Resolution Trust Corporation, or sold to
banks or other financial intermediaries.
Similarly, banks have securitized and sold a large fraction of their assets in recent
years, and the slowdown in bank credit in the table could understate the availability of
bank-originated credit. Moreover, the continued rapid growth of nondepository credit
relative to GDP suggests that much more credit is being held outside the banking system
than in the past as financial innovation has permitted greater access to the capital markets by corporations.
Table 3 contains a breakdown of total bank lending by U.S. chartered banks and foreign banks. The table shows that all categories of bank lending at U.S. chartered banks
grew rapidly in the mid-1980s, and mortgage lending continued strong through the late
1980s. All types of U.S. bank lending slowed sharply in the 1989-91 period, and C & I
lending went into a decline, reflecting not only supply-side restrictions but also the
slower pace of economic activity. However, it appears that some of the reduction in
lending by U.S. banks was picked up by the foreign banks (right side of Table 3). A l though the lending by foreign banks did slow, the rate of increase remained substantial,
suggesting that the foreign banks continued to be active lenders during much of this
7
For alternative reviews of recent credit Hows, see Fred Furlong, "Financial Constraints and Bank Credit,"
Federal Reserve Bank of San Francisco Weekly Letter, May 24, 1991; and Steven Slrongin, "Credit Flows
and the Credit Crunch," Federal Reserve Bank of Chicago, Chicago bed Letter, November 1991; and
Robert Parry, "The Problem of Weak Credit Markets: A Monetary Policymaker's View," Federal Reserve
Bank of San Francisco Weekly Utter, January 3, 1992.
Table 2: Credit Flows (1960-91)
Average Growth Rates
(1)
(2)
(3)
(4)=
(1)-(2)
Depository
Credit
Bank
Credit
(5)
(6)
(8)
(9)
(10)=
(1)-(9)
Bank
Lending
Thrift
Credit
Nondepository
Credit
GDP less
Nondepository
Credit
(7)
Nominal
GDP
Total
Debt
Private
Debt
GDP
less
Total
Debt
1960-79
8.6
8.3
9.5
0.3
9.6
9.3
10.5
10.1
7.2
1.4
1980-82
7.6
9.2
8.2
-1.6
6.3
7.6
6.6
4.0
12.3
-4.7
1983-85
9.1
13.3
12.4
-4.2
11.1
9.9
10.0
13.1
15.3
-6.2
1986-88
6.8
9.8
9.9
-3.9
8.0
7.4
8.9
8.9
11.2
-4.4
1989-91
4.8
6.1
5.1
-1.3
0.1
4.8
4.1
-8.7
10.1
-5.4
Source: Flow of Funds and Commerce Department.
74
credit slowdown period.8 These trends are much the same if the calculations are made
in real rather than nominal terms.
The pace of bank lending has also showed considerable variation across regions,
suggesting that credit restraints were more severe for some regions and less severe for
others (Table 4). The decline in total lending over the 1989-92 period was the sharpest
in the New England, Mid-Atlantic, and Pacific regions, which account for about onehalf of all lending nationwide. In these three regions, C&I lending showed sharp declines, as did real estate and consumer lending. C&I lending declined in all regions during the 89-92 period, while consumer lending declined in all of the regions except two.
Table 4 also shows that residential and commercial real estate lending were not equally
affected. Total real estate lending (commercial and residential) was cut back much more
sharply during the 1989-92 period than residential lending, indicating that commercial
real estate lending was more sharply reduced. In the final section, we will more carefully explore differences in regional lending by looking at regional differences in capital, loan-loss reserves, and employment growth.
Nature and Causes of the Slowdown in Bank Lending
In this section, we explore the slowdown in bank credit from both demand- and supplyside perspectives. We conclude that both sides seem important, a finding that complicates the effort to quantify the causes of the bank lending slowdown.
Demand-Side Explanations
Chart 1 compares depository lending in this recession with lending in past recessions.
In earlier recessions, depository credit grew rapidly until the peak in GDP, and then continued to grow but at a slower pace after the peak. In contrast, not only did depository
lending fail to increase in the period leading up to the most recent recession, but it has
also declined sharply since the peak in business activity.
8
These statistics probably understate the growth of lending by foreign banks because the lending activity
through the offshore offices of foreign banks to U.S. corporations is not adequately captured in the statistics. For more information, sec Robert McCauley and Rama Scth, "Foreign Bank Credit to U.S. Corporations: The Implications of Offshore Loans," Federal Reserve Bank of New York Quarterly Review, Spring
1992, pp. 52-65.
Table 3: Components of Bank Lending
Average Growth Rates
U.S.-Chartered Commercial Banks
Total
Loans
Foreign Banking Offices in U.S.
Mortgages Consumer C&I Loans
Total
Loans
Mortgages C&I Loans
1960-79
9.2
8.8
9.2
9.2
16.9
-
28.1
1980-82
7.2
6.4
0.8
10.7
12.1
65.4
11.1
1983-85
10.3
11.7
15.2
7.3
9.8
6.3
8.3
1986-88
7.5
14.2
7.7
2.8
21.4
48.5
20.5
1989-91
2.9
7.9
0.9
-1.8
15.9
31.5
11.0
75
Causes and Consequences
Table 4: Commercial Bank Lending by Region
Annualized Growth Rates, in Percent
New
England
MidAtlantic
East
North
Central
West
North
Central
South
Atlantic
East
South
Central
West
South
Central Mountain
Pacific
Total loans
1985-87
25.4
11.7
9.4
3.9
16.4
10.4
-6.2
3.2
3.3
1987-89
9.0
8.6
8.6
5.3
12.7
8.1
-8.0
5.4
12.9
1989-92
-13.0
-22.3
0.0
1.7
-0.4
3.2
-1.2
-0.9
-19.3
1985-87
19.6
12.1
13.1
4.0
12.5
12.6
-10.5
0.6
5.3
1987-89
14.0
9.8
10.0
8.7
5.3
5.9
-5.2
-3.1
8.5
1989-92
-21.0
-29.5
-4.7
-2.5
-5.9
-0.5
-6.0
-7.6
-25.4
C&l
Real estate total
1985-87
34.6
22.9
15.3
13.9
25.5
19.8
0.3
9.0
8.9
1987-89
11.5
15.2
13.9
10.5
17.5
14.0
-10.4
1.8
20.9
1989-92
-9.5
-14.8
6.4
10.4
3.8
8.4
2.0
2.7
-14.5
Real estate 1-4 families
1985-87
28.2
19.5
10.6
10.7
27.1
20.3
3.0
9.1
0.6
1987-89
12.4
13.4
11.2
11.3
19.5
15.6
1.2
3.2
24.8
1989-92
-4.5
-6.0
9.1
16.6
10.3
12.3
1.4
10.3
-17.8
Consumer and others
1985-87
-4.9
2.6
4.5
8.5
16.6
6.3
-10.2
5.2
-2.4
1987-89
14.6
6.0
5.3
1.6
14.2
5.8
2.6
20.9
4.7
1989-92
-16.3
-19.0
-7.0
-1.8
1.6
-0.4
2.2
-2.9
-19.3
Percentage of total loans as a national total
1985-87
8.4
27.5
13.9
6.2
13.9
3.8
9.2
3.5
13.5
1987-89
9.5
28.2
14.0
5.8
15.3
3.8
6.7
3.3
13.4
1989-92
9.0
23.7
16.0
6.6
18.0
4.6
6.4
3.6
12.1
New England: Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont;
Mid-Atlantic: New Jersey, New York, Pennsylvania; East North Central: Illinois, Indiana, Michigan,
Ohio, Wisconsin; West North Central: Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota,
South Dakota; South Atlantic: Delaware, District of Columbia, Florida, Georgia, Maryland, North
Carolina, South Carolina, Virginia, West Virginia; East South Central: Alabama, Kentucky, Mississippi,
Tennessee; West South Central: Arkansas, Louisiana, Oklahoma, Texas; Mountain: Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming: Pacific: Alaska, California, Hawaii,
Oregon, Washington.
76
Bank lending (Chart 2) does not show as pronounced a change as depository credit.
It grew more slowly prior to the peak in business activity than it had in past cycles, and
it has remained flat since the peak (as noted earlier, a large part of the slowing of total
depository credit had come from the restructuring of the thrift industry). For perspective, Chart 3 shows a cyclical comparison similar to that in Charts 1 and 2 for nominal
GDP. These three charts indicate that the cyclical differences in depository lending and
bank lending are significantly more pronounced than the cyclical differences in GDP,
suggesting that the weaker lending observed in this recession is not just related to the
performance of GDP.
Of the components of bank lending, consumer loans appear to have followed a pattern quite similar to that of total lending relative to past cycles (Chart 4). Consumer
loans were weaker than average preceding the peak in GDP and have declined somewhat since the peak. Once again, the relative weakness in consumer loans appears greater than is the case for GDP. The wide margins that banks have maintained between their
consumer lending rates and the rates paid on consumer small time deposits appears to
have prompted many consumers to use the proceeds of maturing CDs to pay down credit
card and other bank debt.9 (The effects of the credit slowdown on the liability side of
the banking system's balance sheet is covered in the Hilton-Lown paper contained in this
volume.)
C&I lending at banks has been declining slowly during this entire period (Chart 5),
breaking with the usual pattern of growing fairly rapidly during the period before the
peak in GDP and then slowing, though continuing to grow, after the peak. Comparing
this result to Chart 3, we cannot explain the weakness in C&I loans just by the unusual
behavior of GDP in this cycle. In addition, if we examine C&I loans to the corporate
and noncorporate business sectors (Charts 6 and 7), it appears that the smaller firms in
9
An analysis of the behavior of small time deposits during the credit slowdown and the implications for M2
can be found in John Wenninger and John Partlan, "Small Time Deposits and the Recent Weakness in
M2," Federal Reserve Bank of New York Quarterly Review, Spring 1992.
Chart 1: Depository Lending
Percent
110
77
Causes and Consequences
Chart 2: Bank Lending
Percent
110
Chart 3: Nominal GDP
Percent
110
78
Chart 4: Consumer Loans at Commercial Banks
-
3
-
2
-
1 0 1 2
Quarters relative to peak
Chart 5: C&l Lending
Percent
110
79
Causes and Consequences
Chart 6: C&l Lending to Corporate Business
Chart 7: C&l Lending to Noncorporate Business
Index: Peak=100
115
110
105
100
95
90
85
80
75
80
Peak
the noncorporate sector (which are more reliant on bank credit) may have experienced
a sharper reduction in their access to bank credit since the peak in the business cycle than
did larger incorporated business firms.
Loans to finance inventories appear to have been an important determinant of C&I
lending over the business cycle in the past, and this relationship seems to hold in the
most recent cycle as well (Chart 8). Chart 8 shows that the growth rates of inventories
and C&I loans have been slowing sharply since 1989. However, it is not clear from the
chart whether lower levels of desired inventories are reducing the demand for bank
loans or whether the unavailability of bank loans is forcing firms to operate with less
inventory. Some analysts have argued that better inventory management was an important demand-side determinant of the weakness in C&I lending by banks.10
Now important might this improved inventory management be in explaining the
weakness in C&I lending? In Chart 9, we show a cyclical comparison of the ratio of
business inventories to sales.'' Over this most recent cycle, this ratio has ended up about
7 percentage points below the normal pattern; in other words, the level of nominal inventories is about $75 billion less than would be expected on the basis of the pattern in
past cycles.
It is more difficult to judge how weak business loans are relative to a "normal cyclical pattern." Chart 10 compares C&I loans in the most recent cycle with the average of
past cycles. It suggests that C&I lending is about 37 percentage points weaker, which
amounts to about $288 billion. The $75 billion lower level of inventories is about
25 percent of $288 billion. Thus it appears that better inventory management did have
some impact on C&I lending, but it only explains a relatively small fraction of the overall weakness.
Alternatively, in Chart 11 we compare the ratio of C&I lending to business sales in
the most recent cycle with the ratio for earlier cycles. Lending, relative to the level of
business activity, appears to be about 24 percentage points weaker, or $187 billion less
than in previous cycles. This cyclical comparison suggests that the $75 billion shortfall
in inventories would amount to about 39 percent of the unusual weakness in C&I lending. These orders of magnitude, 25 to 39 percent, are intended to be just rough estimates
of the possible role of inventories from the demand side and could overstate the impact
from the demand side to the extent that banks systematically cut back on business lending during the crunch period, including lending to finance inventories.12
An analysis of the role of inventories in the recent credit slowdown can be found in Kevin Klicscn and
John Tatom, "The Recent Credit Crunch: The Neglected Dimension." Federal Reserve Bank of St. Louis
Review, September/October 1992, pp. 18-36. For evidence of improved inventory management, see Dan
Bechter and Stephen Stanley, "Evidence of Improved Inventory Control," Federal Reserve Bank of Richmond Economic Review, January/February 1992, pp. 3-12.
1
' By indexing each cycle relative to the peak quarter, we can eliminate any longer run trend in inventories
and focus more directly on how businesses have managed inventories in this most recent cycle relative to
earlier cycles. The data released on inventories relative to sales during the laic 1980s and early 1990s
showed a pronounced downward trend, and the increase in this ratio during the recession looked rather
small compared with the increase in past recessions. Initially, it appeared that better inventory management was reducing the demand for C&I loans by a substantial amount. Subsequent revisions to the data
significantly increased the level of inventories relative to sales, and now it appears that inventories were
probably not as important in explaining the weakness in business lending as we originally believed.
12
In addition, not all inventories are financed at banks; some are financed in the commercial paper market or
at finance companies. If we repeat the above exercise but include business lending at finance companies
and nonfinancial commercial paper along with C&I lending at banks, inventories can account for 15 to
29 percent of the cyclically unusual weakness in total short-term business credit.
81
Causes and Consequences
Chart 8: Business Inventories and C&l Loans
Percent
40
Chart 9: Ratio of Inventories to Final Sales
lndex=100
104
82
Chart 10: C&l Loans
lndex=100
140
Chart 11: Ratio of C&l Loans to Final Sales of Business
lndex=100
120
Peak
Source: Commerce Department.
83
Causes and Consequences
Real estate lending, in contrast to C&I lending and consumer lending, appears to
have followed a rather typical cyclical pattern during the most recent cycle (Chart 12).
Although banks have reduced their commercial real estate lending in this cycle, they
have been able to continue to make consumer mortgage loans, unconstrained by the
Regulation Q ceilings or state usury ceilings that often bound them in earlier cycles.
Similar conclusions can be drawn from Charts 13 and 14, which contain cyclical
comparisons for the components of real estate lending. That is, home mortgage lending
has been relatively strong over this most recent business cycle, while other real estate
lending has been flat and considerably weaker than in past cycles (and cannot be explained entirely by the unusual pattern in GDP growth in this cycle).
In sum, the charts in this section suggest the following conclusions: bank lending
has been unusually weak in this most recent cycle and has been decelerating sharply
since the mid-1980s. C&I loans are probably contributing to this trend in two ways from
the demand side. More firms are now able to access the credit markets directly, and better inventory controls may have reduced the demand for bank loans. Consumer loans
have also contributed to the slowing in bank lending, but in this case it appears that consumers are using low-yielding time deposits to pay down high-cost consumer debt. Real
estate lending does not appear unusually weak in total, in large part because in this deregulated banking environment, home mortgage lending has held up quite well relative
to past cycles, even as banks have cut back sharply on their commercial real estate lending.
Supply-Side Considerations
Over the past three years, analysts have pointed to several indicators of a credit crunch
in the banking sector, that is, indicators of supply-side constraints on lending. Two of
the most frequently cited indicators have been the wide spreads between bank lending
rates and bank funding costs in both the corporate and consumer sectors (Charts 15 and
16). Both of these rate spreads are close to or above their previous record levels, suggesting a reduced willingness on the part of banks to lend. In other words, if the weak
credit growth was demand driven, we would expect to see these rate spreads narrow as
banks cut loan rates relative to funding costs to attract new business. Hence, these ratespread charts tend to suggest that the lending slowdown at banks was, at least in part,
supply driven.
Survey results also have been consistent with the notion that banks were less willing
to lend during this period (Charts 17, 18, 19, 20, and 21). Both lenders and borrowers
reported tighter credit standards, particularly in the 1990-91 period. These tighter credit
standards applied to firms of various sizes—small, medium and large—and were applied to C&I loans, commercial real estate loans of all types, and land and development
loans. The responses of the bankers, however, are not very helpful in determining the
start of the credit crunch because the survey was discontinued from 1984 to 1989.
Nonetheless, the percentage of banks reporting tighter standards remained above zero
well into 1991, suggesting that the "credit crunch" was not brief. In most of the charts,
however, the tighter credit standards of 1990-91 do not seem to be the highest recorded
in recent decades; equally tight or tighter standards were in effect in the late 1970s, when
monetary policy was very restrictive.
Banks became more reluctant to lend in part because of increases in charge-off rates
and delinquency rates on all types of bank loans, including consumer loans, C&I loans,
and real estate lending (Chart 22). These problem loans tended to weaken bank capital
positions, and as Chart 23 demonstrates, lending tended to weaken more (or decline
84
Chart 12: Real Estate Loans at Commercial Banks
Chart 13: Home Mortgage Lending at Banks
Percent
Peak
115
110
105
100
95
90
85
80
75
85
Causes and Consequences
Chart 14: Other Real Estate Lending at Banks
Chart 15: Prime Rate Less Federal Funds
Fourth Quarter Moving Average
Percent
86
Chart 16: Spread Between Personal Loan Rates and the 6-Month CD Rate
Chart 17: Borrower Surveys of Credit Availability
Net Percent Reporting "Loans Harder To Get" Than Previous Quarter
87
Causes and Consequences
Chart 18: Lender Surveys of Credit Availability
Net Percent Reporting Firmer Standards than Previous Quarter
Net percent firmer (weighted)
100
Source: Senior Loan Officer Survey, Federal Reserve Board.
1967-77, "Standards for Loan to Nonfinancial Businesses"
1978-83, "Standards to Quality for the Prime Rate"
1990-92, "Standards for Loans to Middle Market Firms"
The Senior Loan Officer Survey results are transformed into an index, "net percent firmer,"
by weighting individual response as follows: considerable firmer (200%), moderately firmer
(100%), unchanged (0%), moderately easier (-100%), and considerably easier (-200%).
Chart 19: Standards of Creditworthiness for C&l Loans
Net percent firmer (weighted)
80
88
Chart 20: Credit Standards for Real Estate Loans
Net percent firmer (weighted)
140
Chart 21: Credit Standards for Land and Development Loans
Net percent firmer (weighted)
120
89
Causes and Consequences
more) over time at those banks with poorer capital positions.13 Again, this evidence is
consistent with a finding that supply-side factors were behind the bank credit slowdown,
and it suggests some possible role for the regulators (discussed later in this section).
13
One should use caution in interpreting this chart as evidence of the seriousness of the crunch at capitalconstrained banks. First, some borrowers did have the ability to switch from weak banks to strong banks
during this period, thereby inflating the strong-bank numbers and reducing the weak-bank numbers with
no real change in aggregate lending. Second, some part of the relatively better performance of the strong
banks slcms from the acquisition of weaker banks, again leading to an overstatement of the impact of their
relative performance on aggregate credit availability. Finally, the distribution of bank assets across capital
classifications is not uniform. Well capitalized banks have just under 65 percent of all assets, adequately
capitalized slightly more than 33 percent, and undercapitalized banks just under 2 percent.
Chart 22: Charge-Off and Delinquency Rates on Loans by Medium and Large Insured
Commercial Banks, by Type of Loan 1982-1992
90
Chart 23: Loans at Weekly Reporting Banks Selected by Capital Status*
Growth from Twelve Months Earlier
Percent
Source: Federal Reserve Board.
Note: Capital Status is determined as of Call Report for 9/30/92. Includes undercapitalized,
significantly undercapitalized, and critically undercapitalized banks.
Chart 24: Security Holdings of Banks as a Percent of Security Holdings of Banks Plus
C&l Loans
Percent
80
91
Causes and Consequences
Chart 24 shows another indicator of the increased reluctance of banks to lend during
this period: banks sharply increased their holdings of securities as a share of C&I loans
plus securities. On the basis of this chart and the comparison contained in Chart 25, we
can conclude that the increased holdings were a fairly typical cyclical response. Although, while the pattern is similar to past cycles, it does appear from Chart 25 that the
increase in security holdings is somewhat more pronounced this time around, perhaps
reflecting normal cyclical developments as well as some of the problems in the banking
industry discussed earlier.
In the "Senior Loan Officer Survey on Bank Lending Practices" Board of Governors
of Federal Reserve System, August 1992, bank loan officers were asked why their banks
had increased securities holdings over the last two and one-half years. Among the fiftynine respondents, thirty-five emphasized that securities offered greater profits, thirteen
mentioned the uncertain economic outlook, eleven claimed that the securities would be
used to fund anticipated increases in loan demand, nine cited a desire to improve their liquidity ratios, and nine gave other reasons. (Banks were allowed more than one answer.)
The motive of greater profits, cited most frequently, probably stemmed in part from the
potential for yield-curve profits (funding longer term securities with short-term deposits)
and the absence of any risk-based capital requirements on government securities.
Governor La Ware, in a statement to Congress, discounted the theory that the Basle
Accord capital standards were the cause of the buildup in security holdings at banks.14
He noted that most banks met the standards and that the increase in holdings of government securities occurred at the stronger banks. He attributed the buildup in government
securities to a move by strong banks, "in an environment of depressed loan demand," to
accommodate deposit inflows by investing in government securities—the same pattern
that occurred at credit unions, financial intermediaries not covered by the Basle Accord.
He also noted that almost half of the acquisitions of securities by banks had been in the
14
John LaWarc before the House Subcommittee on Economic Growth and Credit Formation, April 2, 1993.
Chart 25: Banks' Investments
Percent
130
92
form of government-backed col lateral ized mortgage obligations (CMOs), which have a
maturity that better matches bank liabilities than does the maturity of conventional mortgages. Finally, it appeared to Governor LaWare that the government securities acquired
might be a substitute for state and local government bonds, which had become a less desirable asset for banks since the tax advantages were eliminated in 1986.15
Finally, the statistics on conditions and terms of bank lending can shed some light on
the existence of a credit crunch from the supply side.16 Tables 5 and 6 contain our findings. In general, if banks were more reluctant to lend during the 1989-91 period, we
would expect to see (1) the average size of loans decline as banks became less willing
to finance any given mix of projects, (2) the maturity of loans shorten as banks became
less willing to take longer-term exposures, (3) the spread between the effective loan rate
and the federal funds rate widen as banks required larger spreads to make loans, (4) a
15
16
A more detailed analysis of the increase in the banking system's holdings of government securities can be
found in Jonathan Ncubergcr, "On the Changing Composition of Bank Portfolios," Federal Reserve Bank
of San Francisco Weekly Letter, March 19, 1993.
These statistics are published quarterly as special tables in the Federal Reserve Bulletin.
Table 5: Terms for Short-Term C&l Lending
Loan
Rate
Percent
less Fed Commit- Percent
ment
Funds
Demand
Percent
Floating
Percent
ShortTerm
Percent
Collateralized
28.7
37.3
82.8
40.0
n.a.
28.8
40.0
84.7
44.3
2.3
n.a.
24.6
37.5
83.8
39.8
95
2.3
n.a.
22.6
35.5
84.5
33.3
121.9
74
3.8
n.a.
18.4
32.5
86.6
23.3
1982
194.5
49
2.5
62.0
12.6
25.7
89.6
19.3
1983
193.3
40
1.3
63.2
11.1
29.1
90.1
21.3
1984
190.5
44
1.5
66.8
10.2
31.0
89.4
23.0
1985
199.9
47
1.5
67.0
12.8
22.7
86.1
25.5
1986
267.8
46
1.2
76.7
14.6
25.7
88.1
30.0
1987
292.0
49
1.4
78.6
24.3
30.7
89.1
38.0
1988
303.4
48
1.5
78.2
35.9
36.3
91.7
38.5
1989
288.2
52
1.7
76.2
36.4
40.0
91.1
41.0
1990
303.9
49
1.7
76.8
28.0
33.3
90.1
36.5
1991
286.8
59
2.0
75.0
36.1
37.9
89.3
42.0
1992
282.8
60
1.9
70.1
36.6
38.7
89.0
46.9
1993
(first half)
330.0
54
1.8
68.7
40.3
41.3
89.8
43.1
Year
Average
Size
(Thous.)
Average
Maturity
(Days)
1977
44.0
119
2.5
n.a.
1978
49.5
115
2.4
1979
52.6
113
1980
78.3
1981
93
Causes and Consequences
larger share of loans requiring collateral as banks became more cautious in their lending,
(5) a higher percentage of loans made under commitment as banks reduced discretionary lending and made only those loans they had prior commitments to make, (6) a larger
share of short-term credit extended as demand notes so that banks could call the loans
at the first signs of difficulty, and (7) a larger share of variable rate loans as banks passed
more of the interest rate risk back to their customers. Clearly, some of these indicators
are subject to alternative interpretations and could reflect changes in borrowers1 preferences as well. Our effort, however, in this exercise is only to see whether these statistics,
taken as a whole, are broadly consistent with the notion that lending terms tightened
over the 1989-91 period.
On balance, the evidence from this source is mixed. From 1989 on, the average maturity of short-term loans (loans with maturity of one year or less, about 90 percent of
all loans) tends to lengthen somewhat, not shorten, as might be expected if banks were
becoming more restrictive in lending (Table 5). The spread between the loan rate and
the funds rate does widen, and remains above the 1983-88 period, but does not become
as wide as in the late 1970s and early 1980s. The percentage of loans made under commitment peaks in 1988 and then gradually declines, the same pattern apparent in the per-
Table 6: Terms for Long-Term C&l Lending
94
Long-Term
Average Loan Rate
Maturity
less Fed
(Months)
Funds
Percent
Commitment
Percent
Floating
Percent
Percent Collateralized
Long-Term
Year
Average
Size
(Thous.)
1977
46.2
43
3.1
n.a.
45.7
17.2
60.8
1978
62.0
43
2.3
n.a.
55.2
15.3
66.3
1979
60.9
46
2.1
n.a.
56.6
16.2
61.8
1980
100.6
45
2.8
n.a.
68.0
15.5
56.3
1981
135.2
47
4.4
n.a.
71.5
13.4
50.0
1982
147.5
46
3.5
65.1
66.2
10.4
47.0
1983
141.7
52
2.8
66.3
72.1
9.9
55.0
1984
123.2
46
2.8
76.8
72.6
10.6
49.8
1985
157.1
49
2.8
77.0
67.3
13.9
47.0
1986
211.4
52
2.3
76.7
75.2
11.9
51.8
1987
231.9
49
2.2
74.0
67.6
10.9
52.0
1988
215.5
47
2.4
62.9
67.5
8.3
63.8
1989
223.8
48
2.7
69.2
76.5
8.9
65.5
1990
227.5
45
2.8
74.8
75.6
9.9
67.0
1991
200.9
42
3.3
72.2
75.4
10.7
68.5
1992
172.5
45
3.2
73.7
72.2
11.0
68.0
1993
(first half)
176.0
43
3.3
74.2
68.7
10.2
64.6
centage of loans made as demand credits.17 The percentage of floating rate notes,
however, increases sharply in 1989, as does the percentage of loans requiring collateral;
these results are consistent with tighter lending standards. Finally, the percentage of total loans consisting of short-term loans does not show any break in trend that would suggest a tightening of standards.
For long-term loans, the three series that suggest some tightening in lending terms
are the rate spreads, the percentage of loans made as floating rate loans, and the percentage requiring collateral (Table 6). The other indicators once again are mixed, at best.
Hence, for both short-and long-term lending, there is some evidence of a tightening
of credit standards, but the evidence is not consistent across all the indicators examined.
The most convincing evidence comes from the sharp increase in the percentage of loans
requiring collateral. In 1985, about 25 percent of short-term loans required collateral,
in 1989 41 percent. For long-term loans, the percentage increases from 47 percent to
65.5 percent.
The supply-side indicators reviewed thus far in this section suggest that a credit
crunch did occur in the banking sector over the 1989-91 period. Some analysts have
argued that the regulators could have played an important role in creating the credit
crunch. With several factors from both the demand side and supply side apparently contributing to the slowdown in bank lending, however, the role of the regulators in this
credit slowdown is difficult to determine empirically with any degree of precision, although the regulators certainly seem to have received relatively more attention in the
press. To the extent that the regulators forced banks to face up to the reality of the real
estate market and the state of the economy, they were merely messengers with bad news.
That is, the effects of some of the other factors contributing to the bank credit slowdown
were transmitted through the regulatory process, making the exogenous role of the regulators seem larger than it actually was. Nevertheless, to the extent that regulators became overly aggressive in pressing banks to raise credit standards, they could have been
some part of the reason for the credit crunch. Governor LaWare, in a recent statement
before Congress, provided a summary of the regulatory process during this period.18
Hindsight tells us, he noted, that the regulators should probably have acted sooner and
more vigorously to avert the problems in the banking industry during the boom phase,
but it would have been difficult to impose such standards in good times. Governor LaWare also addressed the question of excessive tightening of credit standards and the regulatory agencies1 responses. He observed that concerns about excessive tightening of
credit standards resulting from the growing volume of real estate lending criticized by
examiners had prompted the regulatory agencies to review their examination practices.
At the same time, however, some tightening of credit standards was necessary given the
state of the economy, and the main concern of the regulators was about excessive tightening resulting from a misunderstanding by banks of supervisory actions. As a result,
the regulators undertook the following actions:
17
Earlier studies using commitment statistics also did not find evidence of rationing. For more detail, see
Allen Berger and Gregory Udell, "Some Evidence on the Empirical Significance of Credit Rationing,"
Journal of Political Economy, 1992, vol. 100, no. 51, pp. 1048-77. In Table 5, the percent of loans made
under commitment had a steady uptrend, peaking in 1987, and slowly declining since then. While higher
on average over the 1989-91 period than the 1982-88 period (76 percent versus 70.3 percent), it is still difficult to point to clear evidence of a credit crunch because of upward trend until 1987.
18
John La Ware, testimony before the Committee on Banking, Finance and Urban Affairs, U.S. House of
Representatives, July 30, 1992.
95
Causes and Consequences
• Accordingly, building on earlier initiatives, in March 1991 the agencies issued a
joint statement to address this matter. That statement sought to encourage banks to
lend to sound borrowers and to work constructively with borrowers experiencing
temporary financial difficulties, provided they did so in a manner consistent with
safe and sound banking practices. The statement also indicated that failing to lend
to sound borrowers can frustrate bank efforts to improve the quality and diversity of
their loan portfolios. Under-capitalized institutions and those with real estate or other asset concentrations were expected to submit plans to improve their positions, but
they could continue sound lending activities provided the lending was consistent
with programs that addressed their underlying problems.
• At other times during the year, and particularly in early November, the agencies expanded on that March statement and issued further guidance regarding the review
and classification of commercial real estate loans. The intent was to ensure that examiners reviewed loans in a consistent, prudent, and balanced fashion. This second
statement emphasized that evaluation of real estate loans should be based not only
on the liquidation value of collateral, but also on a review of the borrower's willingness and ability to repay and on the income-producing capacity of the properties.
• Finally, in December, in order to assure that these policies were properly understood
by examiners and to promote uniformity, the agencies held a joint meeting in Baltimore of senior examiners from throughout the country in one more effort to achieve
the objectives just described. Once again, the principal message was to convey the
importance of balance. Examiners were not to overlook problems, but neither were
they to assume that weak or illiquid markets would remain that way indefinitely
when they evaluated commercial real estate credits.
Because the regulators were simultaneously an independent force as well as a part of
the process through which the banks became aware of the changing economic situation,
their role in the bank credit slowdown cannot be defined precisely. Indeed, other regulators have suggested more strongly than Governor LaWare that the shortcomings of the
regulatory process were not in managing the retrenchment over the 1989-91 period, but
rather in not containing the excesses that were created in the preceding three or four
years.19 That is, they argue that the regulators should have been more aggressive in increasing capital earlier in the 1980s, when the risky lending was actually taking place,
and that this larger capital cushion could have been used to absorb loan losses during the
downturn without cutting off credit to other borrowers. By imposing higher capital standards after the losses became apparent, the regulators, according to Syron and Randall,
may have forced banks to downsize, thereby reducing credit supply to borrowers dependent on intermediated credit. In addition, some banks that were able to meet the riskbased asset requirements by allocating their investments carefully across the various
risk categories were constrained by the higher leverage ratio of tier one capital to unweighted assets imposed by regulators as their condition deteriorated.20 Clearly, invest-
96
19
Richard Syron, "Are We Experiencing a Credit Crunch?" Federal Reserve Bank of Boston New England
Economic Review, July/August 1991, pp. 3-10; and Richard Syron and Richard Randall, "The Procyclical
Application of Bank Capital Requirements," Federal Reserve Bank of Boston, 1991 Annual Report.
20
For more detail and some econometric evidence, see Herbert Bacr and John McElravcy, "Capital Adequacy and the Growth of U.S. Banks," Federal Reserve Bank of Chicago, Working Paper, May 1992. For
evidence from the Boston Federal Reserve District, see Joe Peek and Eric Roscngren, "Crunching the
Recovery: Bank Capital and the Role of Bank Credit, Real Estate and the Credit Crunch, Federal Reserve
Bank of Boston Conference, September 16-18, 1992
ments cannot be reallocated to meet this constraint, and the only option available to
banks in this position is to downsize (that is, if they are unable to raise more tier one
capital, an expensive option for banks experiencing large loan losses). Chairman
Greenspan and Richard Syron have argued that the leverage ratio should be eliminated
as soon as the risk-based measures have been revised to capture the full spectrum of
risks faced by bankers.21 In any case, by the late 1980s, steps were being taken to promote an alternative regulatory approach for banks after the costly savings and loan bailout. Higher capital requirements and insurance premiums were imposed, restrictions on
access to the discount window were established for troubled institutions, and prompt
regulatory intervention for weak institutions was encouraged.22
The slowdown in bank credit that followed these events has been a long, drawn-out
process in part because the economy went through a period of slow growth and recession that lasted three years. Consumers and businesses with heavy debt burdens acquired in the 1980s were in no position to produce a rapid recovery by increasing
spending, even if credit had been readily available. Moreover, fiscal policy, encumbered
by large deficits, could only play a very limited role in turning the economy around.
Monetary policy, in contrast, eased throughout this period, reducing short-term rates
substantially. Borrowers with direct access to the financial markets clearly benefited
from this policy change, but those borrowers that relied on intermediated credit, largely
at banks, may not have benefited as much as they might otherwise have if the banking
system had been in strong financial condition. Hence, the difficulties experienced by the
banking system in recent years may have reduced the effectiveness of monetary policy
by blocking the "credit channel."23
In summary, although analysts have had considerable time to debate the nature and
causes of the "credit crunch," no clear consensus view has emerged even at this late date
as to how much weight to assign to the various forces involved. Because a slowdown
in economic activity lowered credit demand at the same time that (1) consumers and
business worked to reduce heavy debt burdens, (2) banks and regulators lowered their
assessment of real estate values, (3) banks reevaluated their lending standards in light
of the recession, and (4) higher regulatory capital requirements coincided with the
banks' need to increase loss reserves, it has been difficult to sort out the demand and supply causes of the recent credit slowdown. Therefore, while we can find considerable evidence that supply-side factors contributed to the slowdown in bank lending, the
presence of demand-side factors makes it difficult to estimate precisely how much of the
slowdown should be attributed to supply-side considerations. The situation is made
21
For example, sec the transcript of" Chairman Greenspan's statement to the I louse Subcommittee on
Domestic Monetary Policy, Federal Reserve Report to Congress on Monetary Policy, July 22, 1991, pp.
35-36. Also sec statements by John LaFalce, William Taylor, Jerome Powell, and Timothy Ryan, The
Impact of Bank Capital Standards on Credit Availability, U.S. House of Representatives Committee on
Small Business, July 9, 1992.
22
Many of these changes were required by the Federal Deposit Insurance Corporation Act of 1991, December 19, 1991. For additional views of the regulators on the credit crunch, see statements by Alan
Greenspan, Paul Fretts, Robert Clark, and Timothy Ryan in Credit Availability, Committee on Banking,
Housing and Urban Affairs, U.S. Senate, June 2 1 , 1990.
2
^ For more detail, see Donald Morgan, "Arc Bank Loans a Force in Monetary Policy?" Federal Reserve
Bank of Kansas City Economic Review, Q2-1992, pp. 3 1 - 4 1 . An earlier, more theoretical, exposition can
be found in Ben Bcrnanke and Alan Blinder, "Credit, Money and Aggregate Demand," American Economic Review, May 1988, pp. 435-39. A review of the credit channel for a more general audience can be
found in Ben Bcrnanke, "Monetary Policy Transmission Mechanism: Through Money or Credit?" Federal Reserve Bank of Philadelphia Business Review, November/December 1988, pp. 3-11.
97
Causes and Consequences
more complex because other factors, not associated with the excesses of the 1980s, may
also have played a role. As noted earlier, inventories, especially in the manufacturing
sector, may have been managed more closely, and this improvement may have reduced
the demand for short-term business credit by a significant amount.
IV. Econometric Analysis
This section presents an econometric evaluation of the slowdown in bank lending. We
begin with a time series analysis and then offer some cross-sectional results.
Time Series Regressions
To examine the weakness in loan growth in a more formal manner, we present reducedform equations for the four main loan categories. We then compare the actual and predicted values from each equation to assess whether the current weakness in loan growth
is unusual relative to other recessionary periods. We base these comparisons on reduced-form loan equations, as opposed to structural equations, because our interest is
not in determining the true structural model of the loan market. Rather, we seek to establish statistical regularities between loan growth and economic variables that play a
role in determining that growth, and to examine whether those regularities have changed
over the past few years.
The equations are estimated using VAR methodology to approximate the reducedform relationships. Quarterly data from 1961-II through 1991-IV are used in the estimation. The results are presented in Table 7. Each category of loan growth is explained
by four lags of its own growth, four lags of an economic activity variable, and four lags
of an interest rate variable.24 In addition, four lags of investment in producer durables
and in inventory investment are included in the C&I loan equation, four lags of investment in structures are included in the real estate loan equation, and four lags of housing
starts are included in the home mortgage equation. With the exception of the interest
rate and housing start variables, all of the variables are specified as growth rates.
As the table indicates, with perhaps the exception of the C&I loan equation, the regressions appear to explain the loan variables reasonably well. The adjusted R2, one
measure of an equation's fit, is generally between .5 and .6 on a scale of 0 to 1. For C&I
loans, the fit is not as good, perhaps an indication that the nature of business loans has
changed over time. Alternatively, the relatively poor fit of the business loan equation
could suggest that nonprice terms play a role in business lending, or that the many alternatives to bank business loans make it difficult to capture the dynamics of C&I loan
growth.
To determine the stability of our regressions over time, we needed to know whether
any of the regressions would be altered if they were estimated over subperiods of our
sample. First, we split the time period in half, breaking the sample at 1976-HI, and tested whether any of the regressions changed from the first half of the sample period to the
second. Only the C&I loan equation showed a significant change. We then split the
sample period at 1979-1V to determine whether the regressions would differ over the
1980s. This test demonstrated that the home mortgage equation and the C&I loan equation were statistically different during the 1980s than during the 1960s and 1970s. The
tests on the other equations indicated that, for the breaks we considered, our equations
were stable over the sample period.
24
98
Results arc similar when two lags of each variable are used.
Table 7: Bank Loan Equations
Estimation Period: 1961-11 to 1991-IV
Dependent Variable Growth Rate of:
(1)
Explanatory
Variables
C&l Loans
(2)
Real Estate
Loans
(3)
Home Mortgage
Loans
(4)
Consumer Loans
.01
(1.25)
.005
(1.10)
-.007
(1.35)
.002
(.46)
M
.10
(1.1)
.45
(.74)
.45
(4.70)
.53
(5.52)
Lag. dep. var. t_2
.31
(3.4)
.36
(3.46)
.26
(2.47)
.25
(2.25)
Lag. dep. var. t_3
.04
(.37)
-.05
(.44)
-.07
(.68)
.13
(1.18)
Lag. dep. var.
.05
(.54)
-.007
(.06)
.01
(-05)
-.26
(2.78)
it-1
-.003
(1.60)
-.002
(1.84)
-.002
(.88)
-.005
(1.29)
't-2
.002
(1.43)
-.002
(1.62)
-.002
(1.26)
-.003
(.90)
•t-3
.001
(.51)
-.002
(1.57)
-.001
(.49)
-.004
(1.17)
•t-4
.000
(.16)
.002
(1.75)
-.000
(.27)
-.005
(1.59)
XGDP M
-.07
(-36)
.09
(.70)
XGDPt.2
-.37
(1.80)
-.09
(.68)
XGDPt_3
-.06
(.28)
-.06
(.45)
XGDPM
.07
(.34)
.06
(.47)
YPt-1
-.04
(.27)
-.02
(.11)
YPt-2
.01
(-08)
.23
(1.44)
YPt-3
.07
(.48)
-.12
(.77)
YPM
.13
(.86)
-.19
(1.14)
Constant
Lag. dep. var.
M
It-1
.06
(.75)
-.006
(-.16)
"t-2
.02
(.21)
.04
(1.00)
't-3
.05
(.61)
.04
(•99)
lt-4
.28
(3.89)
.03
(.64)
Notes: (absolute) t-statistics are in parentheses. All variables are in nominal terms and are
specified in growth rates except where noted. The real estate loan variable excludes residential
mortgages. XGDP= nominal GDP less expenditures on producer durables, structures, and
inventories. 1= expenditures on producer durables in equation 1 and expenditures on structures
in equation 2. YP= personal income. INV= the stock of business inventories. i= the change in
the prime rate in equations 1 and 2, the change in the mortgage rate in equation 3 and the
change in the consumer loan rate in equation 4. The consumer loan rate is a weighted average
of four consumer loan rates; actual data was available beginning in 1973-1, extrapolated data
was used prior to that. HS= the level of housing starts.
(Continued)
99
Causes and Consequences
Table 7: Bank Loan Equations (Continued)
Estimation Period: 1961-11 to 1991-IV
Dependent Variable Growth Rate of:
(1)
Explanatory
Variables
C&I Loans
INVM
-1.0
(.62)
INVt_2
-.08
(.47)
INVt_3
.28
(1.64)
INVt_4
-.14
(.73)
(2)
Real Estate
Loans
(3)
Home Mortgage
Loans
HSM
-.000
(1.23)
HSt_2
.000
(.98)
HS t . 3
-.000
(.95)
HSM
.000
(2.16)
Adj.R 2
.39
.53
.61
(4)
Consumer Loans
.60
Notes: (absolute) t-statistics are in parentheses. All variables are in nominal terms and are
specified in growth rates except where noted. The real estate loan variable excludes residential mortgages. XGDP= nominal GDP less expenditures on producer durables, structures, and
inventories. 1= expenditures on producer durables in equation 1 and expenditures on structures in equation 2. YP= personal income. INV= the stock of business inventories. i= the
change in the prime rate in equations 1 and 2, the change in the mortgage rate in equation 3
and the change in the consumer loan rate in equation 4. The consumer loan rate is a weighted
average of four consumer loan rates; actual data was available beginning in 1973-1, extrapolated data was used prior to that. HS= the level of housing starts.
Finally, we estimated each equation through 1989-1V and examined the 1990-1 to
1991 -IV out-of-sample prediction errors. This test should reveal whether there has been
a recent structural break. The results are reported in Table 8. The table lists the actual
and predicted values for each of the four loan variables over the eight forecasted quarters. The average error and the root mean squared error are also reported. In addition,
we report the marginal significance level of the F statistic for the test that the prediction
errors are small enough to be consistent with the estimated model.
As the table indicates, the C&I loan equation has a relatively small average prediction error; the errors increase as we move to the real estate, home mortgage, and consumer loan equations. The root-mean squared errors indicate that the consumer loan
equation forecasts somewhat better than the two real estate equations. This latter finding is consistent with the marginal significance levels of the F-tests, which indicate that
only the C&I and consumer loan forecasts cannot reject the hypothesis that the data fit
the estimated equation. We also estimated the C&I loan equation beginning first in
1976-III and then in 1979-1V. The out-of-sample forecasts from these two estimations
were similar to those reported in the table. Thus, while our previous break tests indicated that the C&I loan equation is not stable over the entire sample period, no break occurs
100
over the last two years.25 The two real estate equations have a significant break after
1989-IV.
With the caveats in mind that the behavior of the home mortgage and C&I loan equations changes by the early 1980s and that the two real estate loan equations change after
1989-IV, we next examine the prediction errors from each equation estimated over the
entire sample period. This exercise allows us to assess the current weakness in loan
growth relative to its behavior in other recessionary periods.
The differences between the actual and predicted loan growth values are shown in
Charts 26-29. Chart 26 presents the C&I loan growth prediction errors. As the chart
indicates, the errors in the C&I loan equation do not show any clear cyclical patterns,
and the most recent recession does not show unusually large errors (a result consistent
with the findings reported in Table 8). However, this result should not be interpreted to
mean that a "credit crunch" did not occur for the business sector. To the extent that certain businesses are dependent on bank credit, banks' reduced willingness to lend would
constrain these firms' level of business activity without any errors necessarily showing
up in an aggregate equation.
25
However, there was evidence of a break in the C&I loan equation, significant at the 10 percent level, after
1987-1V.
Table 8: Loan Equations' Forecast Accuracy
Estimation Period: 1961-11 to 1989-IV; Forecast Period: 1990-1 to 1991-IV
Real Estate Loans
(In Percent)
C&I Loans
(In Percent)
Home Mortgages
(In Percent)
Consumer Loans
(In Percent)
Actual
Predicted
Error
Actual
Predicted
Error
Actual
Predicted
Error
Actual
Predicted
Error
1990-1
2.07
1.08
0.99
2.26
2.78
-0.52
3.84
2.78
1.06
0.40
2.35
-1.96
1990-11
2.02
0.82
1.20
3.12
2.43
0.70
3.03
2.06
0.97
-0.20
1.97
-2.17
1990-111
0.64
0.24
0.40
1.93
2.30
-0.37
5.00
2.01
3.00
0.46
2.28
-1.82
1990-1V
0.88
0.59
0.29
1.70
2.25
-0.55
4.03
1.92
2.11
-0.24
2.51
-2.76
1991-1
-1.24
1.04
-2.28
-4.82
2.76
-7.58
10.51
2.09
8.42
-0.65
2.41
-3.06
1991-11
0.09
1.04
-0.95
0.03
2.57
-2.54
4.62
1.82
2.81
-0.96
2.54
-3.50
1991-111
-0.30
1.68
-1.98
-0.96
2.99
-3.95
3.92
1.38
2.54
-2.06
2.28
-4.35
1991-IV
-0.67
-0.29
-0.38
0.05
2.76
-2.71
2.45
1.03
1.41
-1.09
2.57
-3.66
Average
error
-.34
-2.19
2.79
-2.91
Root
mean
square
error
1.27
3.39
3.58
3.03
F-test significance
.85
.00
.00
.20
lOl
Causes and Consequences
Chart 26: Commercial and Industrial Loan Growth Prediction Errors
Percent
8
Chart 27: Real Estate Loan Growth Prediction Errors
102
Chart 28: Residential Mortgage Loan Growth Prediction Errors
Chart 29: Consumer Loan Growth Prediction Errors
103
Causes and Consequences
The prediction errors for the remaining equations are more striking. As Chart 27
shows, the errors for real estate loan growth, excluding home mortgages become huge
in the 1990 recession. The opposite pattern is apparent in the home mortgages equation.
The actual level of home mortgages was much larger than its equation predicts, resulting
in unusually large positive errors during the 1990 recession (Chart 28). This finding also
adds support to the earlier observation that growth in residential mortgages during the
recent recession was stronger than growth in this series during any other postwar recession. This result is also consistent with the earlier findings that this regression is different during the 1980s. The errors from the equations, which are based on past
relationships among economic variables, indicate that during the 1990 recession a
change in the behavior of both real estate lending and home mortgages relative to their
past cyclical behavior seems to have taken place. Problems with commercial real estate
have reduced banks1 willingness to lend to this sector, and commercial real estate lenders
have consequently needed to reduce their reliance on bank loans, leading to the large
negative errors from the equation. At the same time, we see large positive errors in the
home mortgage equation because of deregulation, the downsizing of the thrift industry,
favorable capital treatment of these assets, and the greater liquidity afforded by the ability to securitize these assets as necessary.
The final loan component to consider is consumer loans. Chart 29 indicates that in
previous downturns this loan category showed some weakness that was not predicted by
our equation. But during the 1990 recession our equation consistently overpredicts this
category of loans, no doubt reflecting the wide interest rate spreads between consumer
loan rates and deposit rates that have been prompting consumers to use bank deposits to
pay down bank debt.
We conducted one further exercise to determine more precisely whether the behavior
of lending during the recent recession differed from that in previous recessions. We included in the regressions a dummy variable for each of the recessions in our sample period, as well as one for the 1966 credit crunch. Each dummy variable is equal to one
during the quarters of the recession or credit crunch it is representing, and zero otherwise. 26 The coefficients and their significance for the five variables in each regression
are reported in Table 9.
26
We treated the 1980-81 and 1981-82 recessions as one long recession.
Table 9: Bank Lending During Credit Crunches8
Recession Period
C&l Loans
Real Estate
Loans
Home Mortgage Loans
Consumer
Loans
1966-111 to 1966-1V
0.01
(0.48)
-0.003
(-0.42)
0.000
(0.05)
-0.008
(-1.05)
1969-1V to 1970-1V
0.004
(0.52)
-0.004
(-.85)
-0.01
(-1.61)
-0.005
(-0.89)
1973-1V to 1975-1
0.02
(1.86)
0.003
(0.61)
-0.01
(-1.55)
-0.01
(-2.24)
1980-1 to 1982-IV
0.005
(0.86)
.001
(0.24)
-0.01
(-1.54)
-0.01
(-1.60)
1990-111 to 1991-1
-0.01
(-.85)
-0.03
(-4.46)
0.04
(6.19)
-0.01
(-1.53)
a
- Coefficients and t-statistics (in parentheses) for dummy variables by loan category for each
recession period.
104
The table confirms that the weakness of lending in this recession relative to previous
recessions is unusual. The dummy variable representing the 1990-91 recession is highly
significant in the two real estate equations and significant at the 13 percent level in the
consumer loan equation. The positive coefficient in the home mortgage equation and
the negative coefficients in the real estate and consumer loan equations are consistent
with the respective under- and overprediction that we noted in the residual charts.
Interestingly, in the C & I loan equation the dummy variable for the 1973-75 recession
is positive and significant. This finding suggests that in this earlier episode, the banking
system may have been a greater source of strength to firms having problems funding
themselves in the money market than it was in this most recent period. Home mortgages
were somewhat weaker than predicted over the first three recessionary periods, as evidenced by the 10 to 15 percent significance level of the three dummy variables. This
finding probably reflects the effects of Regulation Q and state usury ceilings. Finally,
the 1966 credit crunch period does not appear to have had an unusual effect on any of
the loan categories.
Overall then, with the possible exception of the C & I loan equation, all of the equations indicate that the behavior of loans has been unusual during the recent recession.
Residential mortgages have been much stronger than their previous behavior would
have suggested, while consumer loans and real estate loans have been considerably
weaker. Conclusions about C & I loans are difficult to draw because of the instability we
found at various break points.
Cross-sectional Regressions
Having established that the behavior of depository lending during the 1990 recession
differed from its behavior in previous recessions, we next consider the role that bank
capital could have played in this change. In an earlier article examining the 1990 credit
crunch, Bernanke and Lown found that banks' capital-to-asset ratios had a positive and
significant link with bank lending during the quarters of the 1990-91 recession.27 They
found that this link held in cross-sectional regressions across states and across banks in
the state of New Jersey. However, they also observed that the coefficients in these regressions were not large enough to justify attributing all of the slowdown in bank lending to weakness in bank capital. This evidence, along with other findings, led the
authors to conclude that weakness in the demand for bank loans clearly played a role in
the lending downturn.
Peek and Rosengren also examined the link between lending and bank capital but
only for New England banks.28 They, too, found that over the period 1990-1 to 1991-I
a capital crunch occurred in New England. They were unwilling to conclude definitively, however, that the capital crunch affected credit availability since they did not take
loan sales into account in their analysis (perhaps loans were still made and then sold).
In this section, we extend the results of these earlier articles by examining the link
between bank capital and loan growth, not only during the year or quarters of the recession, but also prior to it as well. This approach allows us to consider whether bank capital has always been a determinant of lending or only became a determinant during the
27
Ben Bernanke and Cara Lown, "The Credit Crunch," Brookings Papers on Economic Activity, 1992-11. pp.
205-39.
28
Joe Peek and Eric Rosengren, "The Capital Crunch in New England," federal Reserve Bank of Boston
New England Economic Review, May/June 1992, pp. 21-31; and "The Capital Crunch: Neither a Borrower Nor a Lender Be," Federal Reserve Bank of Boston, Working Paper no. 91-4.
105
Causes and Consequences
106
credit crunch period. This latter finding would be consistent with the idea put forth by
Bernanke and Gertler, and later explored in Samolyk, that there is an asymmetric relationship between credit health and bank lending, with the link being stronger in bad
times or in weak regions.29 This finding would also be consistent with the notion that
during the 1990-91 period, banks were cutting back on loans in order to meet capital
standards.
Table 10 contains the regression results. For each year from 1988 to 1991, we estimated cross-sectional (by state) regressions linking loan growth to capital (as a percent
of assets), employment growth, and loan loss reserves (as a percent of loans). The capital and loan loss variables are intended to proxy for differences in the condition of the
banking system across states, while employment growth serves as a proxy for general
economic conditions.30 For each year, we show a regression with capital by itself, another regression with capital and employment, and finally a regression with capital, employment, and loan losses.
For 1988, only employment and loan losses appear significant in determining loan
growth. The capital position of banks does not appear to be a significant factor. The
same is true for 1989. In 1990 and 1991, in contrast, capital is significant when it is included in the regressions by itself and when employment is also included. The significance of capital weakens, however, when loan losses are included, a result that suggests
some interaction or multicollinearity between the two variables. This outcome probably
stems from the fact that the capital ratio is a book-value measure, while the loan-loss
ratio likely reflects more accurately the current market value of the banking system's
portfolio in each state. Hence, once the loan loss measure is included in the regression,
the capital measure provides little additional information.
In general, the regressions provide some evidence of a "capital crunch" (banks being unable to lend because of weak capital positions) in 1990 and 1991 but not in the
earlier two years. In all four years, there is evidence that banks cut back on lending
when they experience greater loan losses. Differences in the state economies (as proxied by employment growth) tend to be an important determinant of bank lending, as
would be expected.31
Finally, to obtain some additional information about the sources of the credit crunch
across regions, we constructed Table 11, which shows ( I ) the regression errors by region, and (2) the contribution made to the predicted values by each of the independent
variables for 1990 (capital, employment, and losses). These results suggest that the
"credit crunch" was greatest for the New England region; indeed, it was even worse than
predicted (the decline in loans exceeded the substantial drop predicted by that regression
equation). The New Hngland region, followed by the Mid-Atlantic region and the West
South Central region, appears to have the weakest capital position to support lending.
29
Ben Bcrnankc and Mark Gcrller, "Agency Costs, Net Worth and Business Fluctuations." American Economic Review vol. 79 (March 1989). pp. 14-31; and Katherinc Samolyk. "A Regional Perspective on the
Credit View," Federal Reserve Hank of Cleveland Economic Review, 1991-11, pp. 7-38.
•*° Loan loss reserves were intended to capture differences across states in the perceived riskiness of loan
portfolios, whereas the capital variable should capture differences in the ability to lend. Ideally, the capital
variable should be the difference from a desired value, but since the desired value is not known, some of
the effect will be captured in the constant term.
31
Using bank level data, Frederick Furlong also found that the sensitivity of bank lending to the capital position of banks increased in 1990 and 1991. See Furlong, "Capital Regulation and Bank Lending," Federal
Reserve Bank of San Francisco Economic Review, no. 3, 1992.
These regions contributed substantially to reduced lending from loan losses, while lending in New England and the Mid-Atlantic regions also posted declines stemming from
weakness in economic activity as proxied by employment.
How much of the slowdown in bank lending from 1989 to 1990 was due to problems
in the banking industry? To answer this question, we applied the 1989-90 cross-sectional regression coefficients from equation 3 in Table 10 to the changes in the explanatory
variables by region between 1988-89 and 1989-90. In other words, we assumed that the
Table 10: Cross-Sectional Regressions8
IV to
IV
Employment
Capital
Loan Losses
R2
1987-88
-0.02
(1)
-0.06
(1.09)
[1.55]
(2)
-0.47
(1.00)
[1.39]
2.86 (0.88)***
[1.28]**
0.15
(3)
-0.20
(0.84)
[0.98]
2.37 (0.74)***
[1.39]*
-3.54 (0.76)*** [0.87]*** 0.40
1988-89
(1)
0.30
(1.25)
[1.44]
(2)
-0.004
(1.22)
[1.54]
1.16
(0.59)**
(3)
-0.25
(0.90)
[1.82]
2.2
(0.47)*** [0.71]*** -7.05 (1.09)*** [1.38]*** 0.48
-0.02
0.04
[0.80]
1989-90
(1)
2.78
(0.95)*** [1.07]***
(2)
1.95
(0.95)**
[1.10]*
1.23
(0.45)***
(3)
1.41
(1.03)
[1.09]
1.21
(0.45)*** [0.53]*** -1.31
(3')b
0.78
(0.74)
[0.88]
1.38
(0.33)*** [0.25]*** -3.40 (0.89)*** [1.34]*** 0.53
0.13
[0.55]**
0.23
(1.04)
[1.92]
0.24
1990-91
(1)
2.69
(0.73)***
[1.09]**
(2)
1.72
(0.89)*
[1.32]
1.06
(0.57)*
[0.73]
(3)
0.81
(0.74)
[0.79]
0.35
(0.49)
[0.69]
0.20
0.24
-3.63 (0.71)***
[0.99]**
0.50
a
Dependent variable is the fourth quarter to fourth quarter growth rate of total loans aggregated to the state level, including the District of Columbia. Capital is equity capital as a percent
of total assets at the beginning of the period. Employment is the fourth quarter to fourth quarter
growth rate of employment. Loan losses are loan loss reserves as a percent of loans at the
beginning of the period. Standard errors are reported in parentheses. Conventional OLS standard errors are in parentheses. Standard errors based on White's correction for heteroskedasticity are in brackets.
b
- This regression excludes data from two states that exert an unusually large influence on the
coefficients in (3).
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
107
Causes and Consequences
Table 11: Decomposition of Loan Growth by Regiona
1989-1V to 1990-1V, in Percent
Loan Growth
Contribution to Loan Growth
Region
Actual
Predicted
Error
Capital
Employment
Losses
Constant
New England
-14.8
-9.0
-5.8
7.2
-5.4
-4.1
-6.7
Mid-Atlantic
0.4
-4.6
5.0
7.7
-1.8
-3.8
-6.7
East North Central
5.1
2.6
2.5
9.6
0.8
-1.0
-6.7
West North Central
4.1
4.4
-0.3
10.7
2.0
-1.5
-6.7
South Atlantic
3.2
2.5
0.7
9.9
0.5
0.5
-6.7
East South Central
4.4
4.2
0.0
10.8
1.3
-1.2
-6.7
West South Central
-2.1
2.0
-4.1
7.8
4.1
-3.2
-6.7
Mountain
1.1
2.4
-1.3
9.1
3.2
-3.2
-6.7
Pacific
12.3
1.6
10.7
8.7
0.7
-1.1
-6.7
a
- In aggregating from the state to the regional levels, each state was weighted by its regional
share of lending.
Table 12: Regional Slowdown in Lending
1989-90 Relative to 1988-893
New
MidEngland Atlantic
East
North
Central
West
North
Central
South
Atlantic
East
South
Central
West
South
Central
Mountain
Pacific
(1) Slowdown in lending
-16.8
-6.9
-1.7
1.5b
-8.9
-3.0
3.1 b
-3.9
-2.4
(2) Due to capital
-1.3
-0.2
0.0
0.5
•0.3
0.1
1.0
0.1
0.8
(3) Due to losses
-3.5
-2.6
-0.2
0.0
-0.2
-0.2
-0.1
-1.3
0.1
(4) Banking (2)+(3)
-4.8
-2.8
-0.2
0.5
-0.5
-0.1
0.9
-1.2
0.9
(5) Percent banking
28.6
40.1
11.8
0.33
5.6
3.3
29.0
30.8
-37.5
(6) Due to employment
-3.2
-2.1
-1.6
-1.2
-1.9
•2.2
1.4
-1.8
-3.6
(7) Percent employment
19.0
30.4
94.1
-80.0
21.3
73.3
45.2
46.2
150.0
a
- Slowdown in lending indicates the change in the growth rate of lending from 1988-89 to 1989-90. Values in
rows 2 and 3, "Due to capital" and "Due to losses," are calculated from the change in the level of the capital-toasset ratio and the loan loss-to-loan ratio, respectively, from the end of 1988 to the end of 1989. Values in Row 6,
"Due to employment," are calculated from the change in employment calculated in the same way as the lending
growth change. The contribution of each variable is obtained by applying the coefficients from the 1989-90
regression (3).
b
- Lending did not slow down in these regions; therefore the statistics do not have the same interpretations as for
the other regions.
108
coefficients on the independent variables estimated across regions could be used in a
rough way to explain the lending slowdown over time in each region.32
The results are shown in Table 12. The New England and Mid-Atlantic regions experienced slowdowns in bank lending that appear to be due to problems in the banking
industry, as did several of the other regions. This outcome naturally invites us to ask
how much of the slowdown in bank lending nationally might have come from problems
in the banking industry. To arrive at a very rough estimate, we weighted the regional
impacts (row 5 in Table 12) by the share of national lending done in each region (Table
4, bottom row). These calculations suggest that about 15 percent (0.7 percentage
points) of the slowdown in lending growth nationally from 1989 to 1990 (4.8 percentage
points) can be explained by supply-side problems in the banking industry, with the remainder coming from changes in economic conditions and other factors. (Employment
explained just over half, 2.5 percentage points, of the slowdown at the national level.)
However, using alternative coefficient estimates, (31 in Table 10) which ignore two
states that appear to be extreme outlyers (Arizona and Nevada), we find that supply-side
problems in banking explain roughly 38 percent of the lending slowdown (1.8 percentage
points). Hence, these results suggest that something on the order of 15 to 40 percent (0.7
to 1.8 percentage points) of the slowdown in bank lending (4.8 percentage points) came
from banking problems.
Conclusions
In this paper, we have reviewed the role of the banking system in the credit slowdown.
We have attempted to show how heavy debt burdens, deregulation and innovation, the
savings and loan crisis, and a series of shocks to the banking system during the 1980s
set the stage for a sharp slowdown in bank lending late in the decade. With the exception
of home mortgage lending, all categories of bank lending were affected in that lending
was weaker than past cyclical experience would have suggested. Moreover, survey results, interest rate spreads, and performance indicators for banks are all consistent with
the notion that supply-side considerations contributed to the slowdown in bank lending.
Regional data are also consistent with this conclusion: lending tended to slow most dramatically in New England and the mid-Atlantic states, the regions where the banking
industry was experiencing the greatest difficulty. Econometric work in the paper's final
section confirms this conclusion.
However, measuring the impact of supply-side considerations is difficult because
several supply- and demand-side considerations appear to have come into play at the
same time, and financial innovations and offshore foreign bank lending have also affected the interpretation of the bank lending statistics. Nonetheless, a very rough estimate
(based on our cross-sectional regressions) would suggest that supply-side factors in the
banking industry account for about 15 to 40 percent of the slowdown in bank lending
from 1989 to 1990.
32
As Bcrnankc and Lown point out, ideally one would like a time scries model of the effect of the various
bank balance sheet variables on bank lending.
109
Causes and Consequences
References
Akhtar, M. Akbar and Betsy Buttrill White. "The U.S. Financial System: A Status Report and a Structural Perspective." In C. Imbriani, P. Roberti, A. Torrisi, eds., // Mercato Unico Del 1992: Deregolamentazione E. Posizionamento Strategico DelV
Industria Bancaria in Europa, Bancaria Editrice S.p.A., Rome 1991, pp. 515-42.
Baer, Herbert and John McElravey. "Capital Adequacy and the Growth of U.S. Banks."
Federal Reserve Bank of Chicago, Working Paper, May 1992.
Bechter, Dan and Steven Stanley. "Evidence of Improved Inventory Control." Federal
Reserve Bank of Richmond Economic Review, January/February 1992, pp. 3-12.
Berger, Allen and Gregory Udell. "Some Evidence on the Empirical Significance of
Credit Rationing." Journal of Political Economy, vol. I (X), no. 51 (1992), pp. 1048-77.
Bernanke, Ben. "Monetary Policy Transmission Mechanism: Through Money or Credit?" Federal Reserve Bank of Philadelphia Business Review, November/December
1988, pp. 3-11.
Bernanke, Ben, and Alan Blinder. "Credit, Money and Aggregate Demand." American
Economic Review, May 1988, pp. 435-39.
Bernanke, Ben, and Mark Gertler. "Agency Costs, Net Worth and Business Fluctuations," American Economic Review, 79, March, 1989, 14-31.
Bernanke, Ben, and Cara Lown. "The Credit Crunch." Brookings Papers on Economic
Activity, 1992:2, pp. 205-39.
Board of Governors of the Federal Reserve System. Report on Credit Availability for
Small Businesses and Small Farms pursuant to Section 477 of the FDIC Improvement Act of 1991.
Browne, Lynn, and Karl Case. "How the Commercial Real Estate Boom Undid the
Banks." Real Estate and the Credit Crunch, Federal Reserve Bank of Boston Conference, September 16-18, 1992.
Friedman, Benjamin. "Changing Effects of Monetary Policy on Real Economic Activity." Federal Reserve Bank of Kansas City, Monetary Policy Issues in the 1990s, August 30-September 1, 1989, p. 70.
Frydl, Edward J. "Overhangs and Hangovers: Coping with the Imbalances of the
1980s." Federal Reserve Bank of New York Annual Report, 1991.
Furlong, Fred. "Financial Constraints and Bank Credit." Federal Reserve Bank of San
Francisco Weekly Letter, May 24, 1991.
. "Capital Regulation and Bank Lending." Federal Reserve Bank of San
Francisco Economic Review 1992 no. 3.
Greenspan, Alan. Statement to the House Subcommittee on Domestic Monetary Policy,
Federal Reserve Report to Congress on Monetary Policy, Wednesday, July 22, 1991,
pp. 35-36.
10
Greenspan, Alan, Paul Fritts, Robert Clarke, and Timothy Ryan. Credit Availablity,
Committee on Banking, Housing and Urban Affairs, U.S. Senate, June 21, 1990.
Jordan, Jerry L. "The Credit Crunch: A Monetarist's Perspective." Federal Reserve
Bank of Chicago's Annual Conference on Bank Structure and Competition, May 7,
1992.
Kane, Edward. "The S&L Mess: What Really Happened?" Boston, MA: MIT Press,
1985.
Kaufman, Henry. "Debt: The Threat to Economic and Financial Stability and Public
Policy," August 27-29, 1986, pp. 15-26.
. Interest Rates, the Markets, and the New Financial World. New York:
Times Books, 1986.
Kliesen, Kevin and John Tatom. "The Recent Credit Crunch: The Neglected Dimension." Federal Reserve Bank of St. Louis Review, September/October 1992, pp. 1836.
LaFalce, John, William Taylor, Jerome Powell, and Timothy Ryan. The Impact of Bank
Capital Standards on Credit Availability, U.S. House of Representatives Committee
on Small Business, July 9, 1992.
La Ware, John. Testimony before the Committee on Banking, Housing and Urban Affairs, U.S. Senate, June 10, 1992.
. Testimony before the Committee on Banking, Finance and Urban Affairs,
U.S. House of Representatives, July 30, 1992.
. Testimony before the House Subcommittee on Economic Growth and
Credit Formation, April 2, 1993.
McCauley, Robert, and Rama Seth. "Foreign Bank Credit to U.S. Corporations: The
Implications of Offshore Loans." Federal Reserve Bank of New York Quarterly Review, Spring 1992, pp. 52-65.
Morgan, Donald. "Are Bank Loans a Force in Monetary Policy?" Federal Reserve
Bank of Kansas City Economic Review, Q2-1992, pp. 31-41.
Neuberger, Jonathan. "On the Changing Composition of Bank Portfolios." Federal Reserve Bank of San Francisco Weekly Letter, March 19, 1993.
Parry, Robert. "The Problem of Weak Credit Markets: A Monetary Policymaker's
View." Federal Reserve Bank of San Francisco Weekly Letter, January 3, 1992.
Peek, Joe, and Eric Rosengren. "The Capital Crunch in New England." Federal Reserve Bank of Boston New England Economic Review, May/June 1992, pp. 21 -31.
. "The Capital Crunch: Neither a Borrower Nor a Lender Be." Federal
Reserve Bank of Boston Working Paper no. 91-4.
. "Crunching the Recovery: Bank Capital and the Role of Bank Credit,"
111
Causes and Consequences
Real Estate and the Credit Crunch, Federal Reserve Bank of Boston Conference,
September 16-18, 1992.
Poterba, James. "Tax Reform and the Housing Market in the Late 1980s: Who Knew
What, and When Did They Know It?" Real Estate and the Credit Crunch, Federal
Reserve Bank of Boston Conference, September 16-18, 1992.
Samolyk, Katherine. "A Regional Perspective on the Credit View." Federal Reserve
Bank of Cleveland Economic Review, 1991 :Q2, pp. 7-38.
Simpson, Thomas. "Developments in the U.S. Financial System Since the Mid-1970s."
Federal Reserve Bulletin, January 1988.
Strongin, Steven. "Credit Flows and the Credit Crunch." Federal Reserve Bank of Chicago, Chicago Fed Letter, November 1991.
Syron, Richard. "Are we Experiencing a Credit Crunch?" Federal Reserve Bank of
Boston New England Economic Review, July/August 1991, pp. 3-10.
Syron, Richard, and Richard Randall. "The Procyclical Application of Bank Capital
Requirements." Federal Reserve Bank of Boston, Annual Report, 1991.
Wenninger, John, and John Partlan. "Small Time Deposits and the Recent Weakness in
M2." Federal Reserve Bank of New York Quarterly Review, Spring 1992.
White, Lawrence. "The S&L Debacle: Public Policy Lessons for Bank and Thrift Regulation." New York, NY: Oxford University Press, 1991.
112
The Link Between the 1980s Credit Boom and
the Recent Bank Credit Slowdown
by Ronald Johnson and Chun K. Lee]
The purpose of this paper is to examine the relationship between the riskiness of a bank's
lending and funding strategies during the credit boom and its lending behavior during
the recent credit slowdown. We must make clear at the outset that we are not testing
whether the most risky banks set off the bank credit slowdown. In fact, it appears that
the least risky (high-capital) banks began to slow their lending growth a few quarters
before the most risky (low-capital) banks began to slow their lending growth (Chart I). 2
Our intention in this paper is only to test whether there is a link between risky bank behavior in the credit boom of the 1980s and loan downsizing in the early 1990s.
Previous research and anecdotal evidence on the recent bank credit slowdown suggest that the banks that cut back most on lending during the recent bank credit slowdown
had the following characteristics: 1) low capital ratios and high exposure to troubled
real estate loans at the beginning of the period; 2) large loan losses during the period;
and 3) high concentration in the Northeast both in terms of the number of banks and the
amount of banking assets.3 The analysis in this paper differs from previous studies in
that it looks beyond these results—to perhaps more fundamental factors—by asking
1
We lhank Akbar Akhtar, Robert McCaulcy, Lawrence Radccki, Richard Cantor, and Anthony Rodrigucs
for helpful comments.
2
Also, several authors have found evidence of a slowdown in lending from nonbank financial institutions.
See Patrick Corcoran, "The Credit Slowdown of 1989-1991: The Role of Supply and Demand," Credit
Markets in Transition, Federal Reserve Bank of Chicago, 1992, pp. 445-62; Mark Carey, Stephen Prowsc,
John Rca, and Gregory Udell, 'The Private Placement Market: Intermediation, Life Insurance Companies
and a Credit Crunch," Credit Markets in Transition, pp. 843-77; Leland Crabbc and Mitchell Post, "The
Effect of SEC Amendments to Rule 2a-7 on the Commercial Paper Market." Board of Governors of the
Federal Reserve System, 1992; and Richard Cantor and Anthony Rodrigucs, "Nonbank Lenders and the
Credit Slowdown," in this volume.
3
See Ronald Johnson, "The Bank Credit 'Crumble'," Federal Reserve Bank of New York Quarterly Review,
Summer 1991, pp. 40-51; Cara Lown and Ben Bcrnanke, "The Credit Crunch," Brookings Papers on Economic Activity, 1992:2, pp. 205-39; Joe Peck and Eric Roscngrcn, "The Capital Crunch in New England,"
Federal Reserve Bank of Boston New England Economic Review, May-June 1992, pp. 21-31; Herbert
Baer and John McElravcy, "Capital Adequacy and the Growth of U.S. Banks," Federal Reserve Bank of
Chicago, Working Paper Series, no. WP-92-11, June 1992; and Thomas Lutton, Regional Analysis of
Bank Lending, CBO Staff Memorandum, Congressional Budget Office, February 1993. Also, for a useful
discussion of much of the recent literature on this subject sec Allen Bcrgcr and Gregory Udell. "Did Risk-
113
Causes and Consequences
Chart 1: Loans at U.S. Banks
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Note: Total loans are a subset of overall bank credit and include commercial and Industrial (C&l), real estate and consumer loans to U.S. addressees. Bank capital position is measured as of fourth-quarter 1989. Bank capital adequacy
is measured relative to the December 3 1 , 1992, total capital requirement of 8 percent. The three stages of bank credit
slowdown are highlighted. The unshaded area is in the credit crunch, the dark shaded area is in the recession and the
light shaded area is in the recovery.
whether banks that exhibited more risky behavior during the credit boom were the ones
that cut back the most during the loan downsizing period.4 In particular, were they:
I) experiencing weaker asset quality; 2) growing loans more quickly than others, particularly commercial real estate loans; and 3) relying more on volatile liabilities.
Footnote 3 continued
based Capital Allocate Bank Credit and Cause a 'Credit Crunch1 In the United States?" Board of Governors of the Federal Reserve System, September 1993.
In past bank credit slowdowns or "credit crunches" the underlying cause was more apparent. This is
because past bank credit slowdowns were the direct result of the process of disintermediation — a reduction in savings deposits at banks that forced banks to reduce their lending. Disintermedialion occurred
when market rates on Treasury bills and commercial paper rose above Regulation Q interest rate ceilings
on savings deposits at banks and thrifts. It is generally agreed that the process of disinlcrmediation was
not a factor in the recent bank credit slowdown, however. For a detailed analysis of previous credit
crunches, sec Albert Wojnilower, "The Central Role of Credit Crunches in Recent Financial History,"
Brookings Papers on Economic Activity, 1980: 2, pp. 277-326.
14
This study shows that there is a link between 1980s balance sheet risks and loan
downsizing in the early 1990s. We also find that the 1980s balance sheets of highly leveraged banks were more risky than the 1980s balance sheets for low leverage banks.
These differences provide a partial explanation for the difference in these groups' lending growth during the early 1990s.
Sample Characteristics and the Data
The Sample
Our analysis is based on a constant sample of U.S. chartered banks that reported data to
the Federal Financial Institutions Examination Council (FFIEC) for the period firstquarter 1985 through third-quarter 1992.5 The sample consists of 9,248 banks with total
assets of $2 trillion (46 percent of total banking industry assets) as of the third-quarter
1992.
The sample, although a subsample of the U.S. banking system, provides a clear picture of the lending behavior of banks that were not involved in merger and acquisition
(M&A) transactions. Banks involved in M&A transactions were excluded from the
sample mainly because there is no straight forward way to separately analyze the leverage strategy and lending growth of the acquiring and target banks over the sample period, first-quarter 1986 through third-quarter 1992.6
We excluded 692 banks from the sample because they purchased other banks through
M&A transactions over the period first-quarter 1989 to third-quarter 1992.7 As of the
third quarter of 1992, this group of bank purchasers held 27 percent ($ 1.2 trillion) of total
banking industry assets. Also, target banks were excluded from the sample because they
did not report data for the entire period. In addition, banks that were formed and later
closed by the Federal Deposit Insurance Corporation (FDIC), as part of a receivership arrangement to liquidate assets and to pay off deposits, were excluded from the sample.
Sample Period
As mentioned in the introduction, this study examines whether there is a link between
bank balance sheet risks in the credit boom and bank lending behavior during the recent
bank credit slowdown. To facilitate the analysis we divide the sample period, first-quarter 1986 to third-quarter 1992, into two nonexhaustive segments.
The first segment, which we call the credit boom, runs from the first quarter of 1986
to the fourth quarter of 1988. By most accounts, this period was characterized by a sharp
5
As of the third quarter of 1992, total banking industry assets were $4.4 trillion. Of this total. U. S. chartered banks held $3.7 trillion and branches and agencies of foreign banks held $701 billion.
6
An important factor motivating M&A activity is bank management's desire to reduce excess capacity and
to achieve efficiencies in banking operations. Sec Larry Radccki, "The Proximate Causes for the Emergence of Excess Capacity in the U.S. Banking System," Federal Reserve Bank of New York. Draft, July
1992, pp. 5-6 and pp. 41-42.
7
Large mergers during the sample period include: Bank of America/Security Pacific (S73 billion in assets);
Chemical/Manufacturers ($66 billion in assets); Nationsbank (NCNBJ/C&S Sovran ($50 billion in assets);
Fleet/Bank of New England ($23 billion in assets); First Union/Southcasi (SI5 billion in assets); Comerica/Manufacturcrs National ($14 billion in assets); Bank One/Valley National (SI 1 billion in assets); Socicty/Amcriirusl ($11 billion in assets); Wachovia/South Carolina National ($7 billion in assets); Norwest/
United Banks of Colorado ($6 billion in assets); ABN Amro/Europcan American ($5 billion in assets).
15
Causes and Consequences
increase in leveraging throughout the economy coupled with a relaxation of lending
standards, especially by banks and thrifts.8 Also, this period is of interest because it
starts in 1986 when Regulation Q interest rate ceilings on savings accounts at banks and
thrifts were abolished and ends in 1988 just prior to the introduction of risk based capital
requirements for banks and the passage of the Financial Institutions, Reform, Recovery,
and Enforcement Act (FIRREA).9
The second segment, which we call the loan downsizing or adjustment period, runs
from the third quarter of 1990 to the third quarter of 1992. This is the period where the
recession-induced effects on credit demand and creditworthiness combine with the
slowdown in bank credit extensions which began in 1989. A cursory examination of the
aggregate data on bank lending growth suggests that the loan downsizing period can be
divided into two parts (Chart 2). In the first phase, loan growth rates decline sharply and
in some cases become negative. In the second phase, loan growth rates begin to recover
as most banks are successful in raising their capital ratios to desired levels and as the
demand for consumer credit starts to show signs of recovery.10
The Data
To measure bank risk-taking activity we employ seven different data series. These data
series were chosen in large measure because they are reasonable indicators of bank lending and funding behavior and because they are available for all the banks in the sample
for the entire sample period.11 These data series cover four facets of bank lending and
funding behavior over the entire sample period. They are: 1) lending growth (the
growth of total, C&I and commercial real estate loans); 2) problem loans (the ratio of
problem loans to total loans); 3) volatile deposits (uninsured time deposits and brokered
deposits); and 4) bank asset size.
C&I loans are loans to businesses operating in the United States. Commercial real
estate loans consists of loans secured by construction, land development, farmland,
multifamily residential properties and nonfarm nonresidential properties. Total loans
are the sum of the three major loan categories (C&I, total real estate (commercial and
residential) and consumer).12
Problem loans are measured as the ratio of troubled loans to total bank credit. Troubled loans consist of loans past due more than 90 days, loans that are nonaccruing, forex
Thc origins, magnitude and implications of the massive buildup in corporate and household debt and credit
risk over the 1980s arc carefully chronicled by Edward Frydl, "Overhangs and Hangovers: Coping with
the Imbalances of the 1980s," Federal Reserve Bank of New York Annual Report, 1991. Also for a
detailed discussion of the building and mortgage credit boom of the 1980s and its aftermath see James Fergus and John Goodman, Jr., 'The 1989-92 Credit Crunch for Real Estate," Federal Reserve System, Board
of Governors, Staff Study no. 164, July 1993.
9
FIRREA, among other things, mandated tighter limits on the amount that thrifts could lend to one borrower
and imposed new capital requirements on thrifts.
10
1
' Our analysis may be biased somewhat in that these variables do not capture the off-balance sheet activities
of large banks.
12
116
The improvement in the capital ratios of U.S. bank holding companies is chronicled by Cantor and
Johnson. That study also analyzes the stock market response to the various methods of improving bank
capital ratios. Sec Richard Cantor and Ronald Johnson, "Bank Capital Ratios, Asset Growth, and the
Stock Market," Federal Reserve Bank of New York Quarterly Review, Autumn 1992. pp. 10-24.
Consumer loans consist of credit cards and related plans, installment loans, single payment loans and all
student loans.
Chart 2: Loans at U.S. Banks
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Note: The three stages of bank credit slowdown are highlighted. The unshaded area is in the credit crunch, the
dark shaded area is in the recession and the light shaded area is in the recovery. Total loans are a subset of overall
bank credit and include commercial and industrial (C&l), real estate and consumer loans to U.S. addressees.
closed properties and restructured loans. Volatile deposits are defined as time deposits
of $ 100 thousand or more and brokered deposits. Bank size is measured in terms of total
balance sheet assets.
Bank balance sheet and income statement data (loans, large deposits, bank capital,
and net income) are taken from the "Call Reports" which are filed with the FFIEC and
collected by the Federal Reserve System. In addition, this study draws on several other
data sources. Bank M&A transactions are taken from the Federal Reserve System's National Information Center (NIC) database on bank structure. State level personal income and employment, which are used to measure regional economic activity, are from
the Bureau of Economic Analysis of the U.S. Department of Commerce.
Some Basic Relationships
This section of the paper focuses on the basic relationship between bank leverage and
balance sheet riskiness in the mid to late 1980s and loan growth during the loan downsizing period. The first part of this section sorts banks into two groups—highly leveraged and low leverage—based on the size of a bank's capital ratio at the end of 1988 and
117
Causes and Consequences
discusses the characteristics of these two bank groups. The second part of this section
presents evidence that highly leveraged banks were the ones that cut back the most on
their lending during the 1990 loan downsizing period. This finding is consistent with
the results found in other studies on the recent bank credit slowdown. The third part of
this section presents evidence that suggests bank leverage and balance sheet riskiness
were linked during the credit boom—highly leveraged banks had more risky balance
sheets than low leverage banks.
Banks Grouped by End-1988 Leverage Ratios
As mentioned above, we divide our bank sample into two groups based on the size of a
bank's capital ratio relative to the average capital ratio for all banks at the end of 1988. l3
Banks with above average capital ratios at the end of 1988 are classified as low leverage
banks, while banks with below average capital ratios at the end of 1988 are classified as
highly leveraged banks.
The distinction between these two groups is quite straightforward. Low leverage
banks are those banks that were able to expand their assets during the credit boom while
at the same time they were able to maintain or to achieve a higher than average capital
ratio. Highly leveraged banks, on the other hand, are those banks that, for whatever reason, did not prevent their capital ratios from lagging behind the all-bank average.
The average capital ratio for low leverage banks (banks that had above average capital
ratios) as of the end of 1988 was almost 9 percent (Table 1). The average capital ratio for
highly leveraged banks (banks that had below average capital ratios) at the end of 1988
was 5.5 percent. The average capital ratio for all banks at the end of 1988 was 6.7 percent.
To determine whether the leverage characteristics of the two bank groups differed
over the sample period we employ the analysis of variance (ANOVA) procedure. The
13
The average capital ratio for all banks in our sample is determined as the ratio of total bank capital to total
bank assets.
Table 1: Average Bank Capital Ratios
All Banks
Highly Leveraged
Banks
Low Leverage
Banks
1985
6.7
5.6
8.6
1986
6.7
5.6
8.6
1987
6.5
5.3
8.7
1988
6.7
5.5
8.8
1989
6.5
5.2
8.8
1990
6.7
5.5
8.7
1991
6.9
5.8
8.7
1992
7.4
6.3
9.1
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Note: Data are for the fourth quarter of each year, except for 1992 when the data are for the third
quarter. The average capital ratio is defined as the ratio of total bank capital to total bank assets.
US
ANOVA procedure is a statistical technique designed to determine whether or not a particular classification of the data is meaningful. This procedure consists of two steps.
In the first step, we test whether the variances of the capital ratios for the two bank
samples were equal over the same period. This null hypothesis was rejected at the 99
percent level of confidence. The F-statistic was 2.9 with 229,647 and 57,039 degrees of
freedom for low leverage and highly leveraged banks, respectively.
In the second step, we test whether the means of the capital ratios for the two bank
groups are equal given that the variances of the two groups are not equal. This null hypothesis was also rejected at the 99 percent level of confidence. The t-statistic was 310.
Highly leveraged banks account for a significant share of banking assets (Table 2).
As of the third quarter of 1992, the highly leveraged bank group held 62 percent ($1.3
trillion) of the assets of the banks in our sample. The bulk of these assets ($1 trillion or
51 percent of the assets in our sample as of the third-quarter of 1992) were held by just
7 percent of the more than 1,800 highly leveraged banks.
Bank Leverage and Loan Downsizing in the Early 1990s
The evidence is clear that there is a relationship between bank leverage at the end of the
credit boom and loan downsizing in the early 1990s (Chart 3). We find that highly leveraged banks were the most aggressive in loan downsizing during the bank credit slowdown. We also find that the growth in lending from highly leveraged banks exceeded
the growth in lending from low leverage banks during the credit boom.
It is difficult, however, to isolate the separate effects on lending growth of bank leverage from the demand-side shocks (the regional economic "luck of the draw") in the
market for bank credit and changes in capital adequacy requirements. This is because
the probability that a bank will engage in loan downsizing depends to some extent on
all three factors. During the loan downsizing period all three of these factors were interrelated as highly leveraged banks bulked quite large in the regions that suffered the most
from the economic slowdown—the Northeast and the West—and among the banks with
Table 2: Basic Characteristics of the Commercial Bank Sample
As of 1992-111
Number of Banks
Assets in Billions
of Dollars
(Figures in parentheses
represent the percent of
the total number of banks)
(Figures in parentheses
represent the percent of
the total number of banks)
9,248(100)
2,026(100)
1,840(19)
1,260(62)
Banks with assets above $1 billion
134(1)
1,041 (51)
Banks with assets below $1 billion
1,706(18)
218(11)
Low leverage banks
7,408 (81)
766 (38)
Banks with assets above $1 billion
59(1)
191 (9)
Banks with assets below $1 billion
7,349 (80)
575 (29)
Bank Classification
Total
Highly leveraged banks
Source: Federal Financial Institutions Examination Council, Reports of Condition.
119
Causes and Consequences
Chart 3: Growth in Loans
Fourth Quarter Growth Rate
Percent
25
~ 1 ° I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
1986
1 9 8 7 1 9 8 8 1 9 8 9 1 9 9 0 1991
1 9 9 2
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Notes: Commercial bank loans are a subset of total commercial bank credit and include C&l,
real estate and consumer loans to U.S. addressees.
The banks are sorted based on the size of their capital ratio relative to the average capital
ratio for all banks as of the fourth quarter of 1988. Low leverage banks have capital ratios that
exceed the all-bank average and highly leveraged banks have capital ratios that were below
the all-bank average.
inadequate regulatory capital ratios.
Although the effects of the regional luck of the draw and capital adequacy can not be
easily separated from the effects of bank leverage, we can, at least, get an idea of the
120
Table 3: A Comparison of the Growth in Total Loans from Highly Leveraged and
Low Leverage Banks by Region and Capital Adequacy
1990-111 to 1992-11
Highly Leveraged Banks
Low Leverage Banks
(Percent)
(Percent)
-4.7
6.2
Northeast
-9.4
-1.7
Other
0.3
8.2
Undercapitalized
-9.8
-32.4
Adequately capitalized
-0.1
6.8
Undercapitalized
-15.8
-4.6
Adequately capitalized
-3.3
6.3
Bank Group
AH banks
Regional groups
Risk-based capital groups
Regulatory leverage groups
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Note: Risk-based capital groups are divided in the following ways: 1) undercapitalized if sum
of tier 1 and tier 2 risk-based capital ratio is less than 8 percent; 2) adequately capitalized if
greater or equal to 8 percent. Regulatory leverage groups are divided in the following ways:
1) undercapitalized if leverage ratio is less than 4 percent; 2) adequately capitalized if greater
or equal to 4 percent.
extent to which bank leverage may have intensified the effects of demand- and supplyside shocks in the market for bank credit. To do this we first control for the region
(Northeast or other) in which the bank operates and compare the loan growth from our
two bank groups. As Table 3 shows, highly leveraged banks posted lower loan growth
regardless of region. This finding holds true, with only one exception, when we control
for capital adequacy.14
In summary, there is evidence of a link between bank leverage at the end of the credit
boom and loan downsizing in the early 1990s. Highly leveraged banks were the most
aggressive in reducing their lending during the loan downsizing period. This is the case
even when we control for region and capital adequacy.
A question arises, however, as to the mechanism that connects a bank's leverage position at the end of the credit boom to loan growth during the loan downsizing period.
In the literature on the recent bank credit slowdown a veritable cottage industry has developed that suggests that the connection runs mainly through the luck of the draw and
capital adequacy effects.15 Our results in this part of the paper suggest that these effects
do not explain completely the apparent link between a bank's leverage at the end of the
credit boom and its loan growth during the loan downsizing period.
14
The one exception involves low leverage banks that were undercapitalized based on the 1992 risk-based
capital standards. This group consisted of a relatively insignificant number of banks (twenty-three) that
held about 1 percent of low leverage bank assets.
15
For a review of the various hypotheses that have been tested in the recent literature on this subject sec Allen
Berger and Gregory Udell, "Did Risk-based Capital Allocate Bank Credit and Cause a 'Credit Crunch1 In
the United States?" Board of Governors of the Federal Reserve System, September 1993.
121
Causes and Consequences
Bank Leverage and Balance Sheet Risks in the Credit Boom
In this part of the paper we examine whether bank leverage and bank balance sheet risks
were correlated during the credit boom. If they were correlated, then bank balance sheet
riskiness during the credit boom may provide the missing link or channel between a
bank's 1988 leverage position and its early 1990s loan growth. That is, a bank's loan
growth in the early 1990s may reflect the credit and financial risks inherent in the mix
of assets, liabilities, and off-balance sheet exposures that resulted from its operations
during the credit boom.
To a large extent, the willingness and ability of banks to raise the riskiness of their
balance sheets can be traced back to the deregulation, financial innovation and leveraging boom that took place during the decade of the 1980s. Regardless of the motivation
for the increase, there is a question of whether bank leverage and the riskiness of bank
lending and funding were linked during the credit boom. That is, did the highly leveraged banks have the most risk laden balance sheets during the credit boom.
We measure balance sheet risk in three ways: 1) loan quality; 2) credit risk; and 3)
volatile deposits. These three measures are not exhaustive; for example, we do not attempt to measure interest rate or foreign exchange risk. When pieced together, however,
these measures give, at least, an impressionistic sense of the overall riskiness of a bank's
balance sheet.
To proxy loan quality, we use the problem loan ratio (the ratio of problem loans plus
foreclosed properties and restructured loans to total commercial bank credit). We employ two credit risk indicators—the ratios of commercial real estate loans and C & I loans
to total loans. Volatile deposits are measured as the ratio of large time deposits plus brokered deposits to total deposits.
The evidence based on the above risk indicators suggests that during the credit boom
bank leverage and balance sheet riskiness were positively correlated. Chart 4, which
presents the problem loan ratio for the two bank groups, shows that the highly leveraged
banks had a much higher ratio of problem loans than did the low leverage banks. This
was the case for the credit boom and for all of the bank credit slowdown, with the gap
widening during the loan downsizing period.
When we examine the connection between bank leverage strategy and the commercial real estate loan ratio we also find that highly leveraged banks held a considerably
larger share of their loans in the form of commercial real estate loans than did low leverage banks. The connection between bank leverage strategy and the share of C & I loans
on a bank's books appears to be consistent with expectations. Highly leveraged banks
held a greater concentration of loans to businesses than did low leverage banks. It is interesting, however, that the share of C & I loans in the portfolios for both bank groups
have declined steadily throughout the credit boom and the bank credit slowdown.
It also appears to be the case that there is a connection between the importance of volatile time deposits as a funding source and bank leverage (Chart 5). By a wide margin,
highly leveraged banks have relied more heavily on volatile deposits for funding than
low leverage banks have. The gap between the volatile deposit ratios for the two bank
groups, as would be expected, declined considerably during the loan downsizing period.
In summary, there appears to be a link between bank balance sheet riskiness and bank
leverage—the group of banks with the most risky balance sheets were the most highly
leveraged during the credit boom. This relationship also holds throughout the bank
credit slowdown. This evidence suggests that our bank sample can be segmented into
two groups: 1) highly leveraged banks with risky balance sheets; and 2) low leverage
banks with lower risk balance sheets.
122
Chart 4: Ratio of Loans
Fourth Quarter Growth Rate
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Note: Commercial and industrial loans is loans to U.S. addressees. The banks are sorted based on the size of
their capital ratio relative to the average capital ratio for all banks as of the fourth quarter 1988. Low leverage
banks have capital ratios that exceed the all-bank average and highly leveraged banks have capital ratios that
were below the all-bank average.
It is quite likely that bank balance sheet riskiness during the credit boom is the missing link that connects 1988 bank leverage to loan growth in the early 1990s. This is because, by most accounts, many banks relaxed their lending standards during the 1980s
and increased their holdings, both in number and dollar volume, of loans that were based
mainly on anticipated growth in cash flows and asset price appreciation. Therefore, it
should not come as a surprise that banks with risk-laden balance sheets found it necessary to downsize in the face of the downturn in economic activity and the collapse in
commercial real estate prices and leveraged buyouts in the late 1980s and early 1990s.
Bank Balance Sheet Risks in the Credit Boom and Loan Growth in the Early 1990s
This section reports the results of our preliminary examination of the connection between risky bank balance sheets in the credit boom and loan downsizing in the early
1990s. The first part of this section discusses the basic methodology that is used in our
analysis. The second part of this section provides evidence of a link between bank bal-
123
Causes and Consequences
Chart 5: Ratio of Large Time Deposits to Total Deposits at Commercial Banks
Percent
18 i -
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Note: Large time deposits are deposits of $100,000 or more. The banks are sorted based
on the size of their capital ratio relative to the average capital ratio for all banks as of the
fourth quarter of 1988. Low leverage banks have capital ratios that exceed the all-bank average and highly leveraged banks have capital ratios that were below the all-bank average.
ance sheet riskiness in the mid to late 1980s and loan downsizing in the early 1990s.
In addition, this part of the paper examines whether the lending behavior of the two
bank groups during the early 1990s differed significantly in response to the credit boom
and contemporaneous factors. The last part of this section compares the relative importance of 1980s balance sheet risks in determining loan growth during the loan
downsizing period.
Basic Methodology
Our analysis is based on cross-sectional regressions of individual bank level data that
are pooled into the two bank groups (highly leveraged and low leverage). The percent
change in total bank loans over the loan downsizing period, third-quarter 1990 to thirdquarter 1992, is the dependent variable. The change in total bank loans is regressed
against four categories of independent or explanatory variables: 1) basic factors; 2)
1980s balance sheet risks; 3) 1980s loan growth; and 4) contemporaneous factors.
The basic factors are the bank's fourth-quarter 1988 capital ratio and bank size.
Based on previous studies, we expect to find a positive relationship between the 1988
bank capital ratio and the growth in bank lending.16 Bank size is measured as the loga-
16
124
See, for example, Ben Bcrnanke and Cara Lown, "The Credit Crunch," Brookings Papers on Economic
Activity, 1992:2, pp. 205-39; Joe Peek and Eric Rosengren, 'The Capital Crunch: Neither a Borrower nor
Lender Be," Federal Reserve Bank of Boston, Working Paper 91-4, February 1991; and Diana Hancock
and James Wilcox, "Capital Crunch or Just Another Recession?" Bank Structure and Competition, Federal
Reserve Bank of Chicago, 1992.
rithm of bank assets and is included to capture differences in lending behavior between
large and small banks.
The 1980s balance sheet risks are the end-1988 problem loan, real estate loan, C&I
loan, and volatile deposits ratios. These variables were described earlier. We expect to
find a negative relationship between these variables and the growth in bank lending in
the early 1990s.
The 1980s loan growth variables are the growth in total, C&I and commercial real
estate loans over the credit boom. We include the 1980s loan growth variables in our
regressions to test whether the loan downsizing in the 1990s resulted because the pace
at which banking assets grew in the 1980s was simply too fast. As a result, the basic
hypothesis is that 1980s loan growth rates should be negatively correlated with loan
growth over the loan downsizing period.
The contemporaneous factors include: 1) a regional dummy that segments the sample into Northeast and other; 2) a capital adequacy dummy that sorts the sample into adequately capitalized—at least 4 percent tier 1 capital, 8 percent total risk-based capital
and 4 percent leverage ratios—and inadequately capitalized groups as of the third quarter of 1990; and 3) a dummy variable that sorts the sample based on whether the bank
was profitable over the loan downsizing period. The tested hypotheses for these variables vary.
The regional dummy is set equal to one in our regression equations if the data on the
dependent variable is for a bank that is in the Northeast and is set equal to zero otherwise. Consequently, the regression coefficient for the regional dummy is expected to be
negative. The capital adequacy dummy is set equal to one if the bank is adequately capitalized as of the third quarter of 1990 and is set equal to zero otherwise. Consequently,
the regression coefficient for this dummy variable is expected to be positive. The bank
profitability dummy is set equal to one if the bank was profitable over the loan downsizing period and is set to zero otherwise. The expected sign of the regression coefficient
for this dummy variable is positive.
1980s Balance Sheet Risks and Loan Downsizing in the Early 1990s
In this part of the paper, we examine the relationship between bank balance sheet risks
from the 1980s and bank lending growth in the early 1990s. The analysis is performed
on a step-by-step basis where we first estimate the relationship between 1990s bank loan
growth and the basic factors (the first level regression). We next estimate the relationship between 1990s bank loan growth and the basic factors, as well as the 1980s balance
sheet risks and loan growth rates (the second level regression). We then estimate the full
regression model including the contemporaneous factors.
Table 4 presents the results from the first level regressions for the highly leveraged
and the low leverage banks. For the highly leveraged banks the 1988 capital ratio variable is not significant in the first level or basic regression equation. This is a somewhat
surprising result given that this variable is significant and positive for low leverage
banks. For both groups the bank size variable is found to be negative and significant.
This result suggests that larger banks reduced their lending by more than smaller banks
during the loan downsizing period.
The results from the second level regressions for the two bank groups are shown in
Table 5. We find that for the highly leveraged banks, two of the 1980s balance sheet
risks—the problem loan and the real estate loan ratios—were significant and negative.
For the low leverage banks the problem loan ratio was the only 1980s balance sheet risk
variable that was significant.
125
Causes and Consequences
Table 4: First Level Regression
Explanatory Variables
Constant term
1988 Capital ratio
Bank size
Highly Leveraged Banks
Low Leverage Banks
(Numbers in parentheses
are t-statistics)
(Numbers in parentheses
are t-statistics)
41.2
(6.8) ***
52.4
(10.9)***
0.8
(1.3)
0.3
(2.0) **
-3.5
(-8.3) ***
4.1
(-9.9) ***
Note: Dependent variable is the third-quarter 1990 to third-quarter 1992 growth rate of total/C&l
loans. Results corrected for heteroskedasticity.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
For the highly leveraged banks, two of the 1980s loan growth variables were significant while none of these variables were significant for the low leverage banks. It is a
bit surprising that in the highly leveraged banks regression the significant 1990s loan
growth variables are both positively related to 1990s loan growth. Despite the fact that
the size of the coefficients is small and that the coefficients are only marginally significant, this result is not consistent with the relationship that was expected.
The results for the full regression models are presented in Table 6. The introduction
of the contemporaneous factors has altered some of the findings from the first and second stage regressions. The most notable change is that the 1988 capital ratio is no longer
significant for the low leverage banks.
As it turns out, the regional economic dummy is only significant for the low leverage
banks. This is not surprising given the high concentration of highly leveraged banks in
the Northeast. The capital adequacy dummy is only significant for the highly leveraged
banks. This is also not surprising given the concentration of banks with inadequate regulatory capital among the highly leveraged banks. The bank profitability dummy is significant and positive in both bank regression equations.
Given that our findings in earlier sections of this paper indicated that the lending behavior of the two bank groups differed in both the credit boom and the loan downsizing
period, we estimated separate bank level regressions for each of the bank groups. As
Table 6 shows, the explanatory power of the full regression model, as measured by the
goodness of fit (adjusted R2), is considerably higher for the highly leveraged banks than
for the low leverage banks. The table also reports the results of a more formal test (the
F-test or Chow test) that also finds that the separate regressions for the two bank groups
are different.
In summary, our findings indicate that there is a direct link between 1980s balance
sheet risks and loan cutbacks in the early 1990s. At the level of the full regression model
for both bank groups, the 1988 bank capital ratio is not significant. Also, the results
from the regression models highlight some of the reasons why the 1990s loan growth
differed across the two bank groups.
126
The Relative Importance of the Link Between 1980s Balance Sheet Risks and 1990s
Loan Growth
In this part of the paper we examine the marginal contribution of bank characteristics in
the full regressions reported in Table 6. The calculation of marginal contributions involve decomposing each independent variable in the full regression model and quanti-
Table 5: Second Level Regression
Highly Leveraged Banks
Low Leverage Banks
(Numbers in parentheses
are t-statistics)
(Numbers in parentheses
are t-statistics)
58.0
(7.7) ***
55.8
(10.5)***
-0.4
(-0.7)
0.3
(1.7)*
-3.6
(-6.9)***
-4.2
(-8.7) ***
Problem loan ratio
-0.4
(-3.9)***
-0.2
(-2.7)***
Real estate loan ratio
-0.2
(-3.4)***
-0.0
(-0.1)
C&l loan ratio
0.0
(0.2)
0.0
(0.1)
Ratio of large time deposits
to total deposits
-0.1
(-1.5)
-0.1
(-1.3)
Percent change in total loans
during the credit boom period
0.001
(2.0) **
0.0
(1.2)
Percent change in C&l loans
during the credit boom period
0.001
(1.8)*
0.0
(0.3)
Percent change in real estate
loans during the credit boom
period
-0.0
(-1.6)
0.0
(0.4)
Explanatory Variables
Constant term
Basic factors
1988 capital ratio
Bank size
1980s balance sheet risks
1980s loan growth
R-square
0.05
0.02
Adjusted R-square
0.05
0.02
Number of observations
1,840
7,408
Note: Dependent variable is the third-quarter 1990 to third-quarter 1992 growth rate of
total/C&l loans. Results corrected for heteroskedasticity.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
127
Causes and Consequences
Table 6: Full Regression Model
Highly Leveraged Banks
Low Leverage Banks
(Numbers in parentheses
are t-statistics)
(Numbers in parentheses
are t-statistics)
56.9
(6.7) ***
82.2
(3.4) ***
-0.9
(-1.4)
0.2
(1.5)
-4.5
(-8.0) ***
-4.9
(-9.5) ***
Problem loan ratio
-0.3
(-3.7) ***
-0.2
(-2.7) ***
Real estate loan ratio
-0.2
(-3.5) ***
0.0
(-0.2)
C&l loan ratio
0.0
(0.2)
0.0
(0.1)
Ratio of large time deposits
to total deposits
0.0
(-1.3)
-0.1
(-1.3)
Percent change in total loans during
the credit boom period
0.0
(1.1)
0.001
(1.7)*
Percent change in C&l loans during
the credit boom period
0.001
(1.7)*
0.0
(0.3)
Percent change in real estate loans
during the credit boom period
-0.0
(-0.3)
0.0
(0.4)
Dummy variable for
Northeast region
-2.9
(-1.2)
-3.0
(-2.9) ***
Dummy variable for
capital adequacy
8.6
(1.9)*
-20.5
(-0.9)
Dummy variable for
banks' profitability
10.2
(7.2) ***
4.4
(5.7) ***
0.09
0.08
1,840
0.03
0.03
7,408
F-test:
F=22.5
F=12.2
Test for difference between
highly and less leveraged banks
F=2.8
Critical F=1.8 @ 5%
Reject the null that the two
regressions are the same
Explanatory Variables
Constant term
Basic factors
1988 Capital ratio
Bank size
1980s balance sheet risks
1980s loan growth
Contemporaneous factors
R-square
Adjusted R-square
Number of observations
Note: Dependent variable is the third-quarter 1990 to third-quarter 1992 growth rate of
total/C&l loans. Results corrected for heteroskedasticity.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
128
fying the added impact of each of these variables on the explained variation or R2 of the
regression model. It must be emphasized that we are discussing marginal and not total
contributions. Unless the explanatory variables are completely independent of each other, the marginal contribution will fall short of the total contribution.
Since our focus is on the marginal contribution of the independent variables and not
on a comparison of the two bank groups, we focus our discussion only on the results
from the highly leveraged banks regressions. The evidence suggests that bank size provided the largest contribution to the explained variation of the model (Table 7). Bank
Table 7: Partial Contribution
Percent
Highly Leveraged Banks
Low Leverage Banks
(Numbers in parentheses
are t-statistics)
(Numbers in parentheses
are t-statistics)
1988 Capital ratio
0.9
1.9
Bank size
42.4
55.6
1980s balance sheet risks
15.8
8.3
Problem loan ratio
8.4
4.9
Real estate loan ratio
6.6
0.0
C&l loan ratio
0.0
0.0
Ratio of large time deposits
to total deposits
0.8
3.4
1980s loan growth
0.8
3.4
Percent change in total loans
during the credit boom period
0.1
1.1
Percent change in C&l loans
during the credit boom period
0.1
0.0
Percent change in real estate
loans during the credit boom
period
0.6
2.3
Contemporaneous factors
34.9
17.7
Dummy variable for
Northeast region
0.9
1.5
Dummy variable for
capital adequacy
1.9
1.1
Dummy variable for
banks' profitability
32.1
15.0
Total
94.9
86.8
Explanatory Variables
Basic factors
Note: Dependent variable is the third-quarter 1990 to third-quarter 1992 growth rate of
total/C&l loans. Results corrected for heteroskedasticity.
129
Causes and Consequences
size accounts for 42 percent of the model's explained variation.17 The second most important explanatory variable is the bank profitability dummy, which accounts for
32 percent of the explained variation of the model.
The 1980s balance sheet risks provided the third largest overall contribution to the
explained variation of the model. Of this group, the problem loan ratio (8.4 percent)
was the most important. It should be noted, however, that by themselves the problem
loan and real estate loan ratios provide the third and fourth largest contributions to the
explained variation of the model.
Conclusion
The task set for this study was to examine whether there is a link between bank balance
sheet riskiness in the 1980s and loan downsizing in the early 1990s. The available evidence suggests that there is a significant relationship between these two measures of
bank behavior. This result should not come as a surprise given the size of the bad loan
hangover from the 1980s.
The most important 1980s balance sheet risk variable was the loan loss reserve ratio.
For the highly leveraged banks, the real estate loan ratio also played an important role.
Overall the 1980s balance sheet risks were more important for the highly leveraged
banks than the low leverage banks in explaining 1990s loan growth.
17
130
This result is consistent with the finding by Boyd and Gcrtler that bank size is a good indicator of a bank's
susceptibility to distress. See John Boyd and Mark Gcrllcr. "U.S. Commercial Banking: Trends, Cycles,
and Policy," New York University, February 1993.
Loan Sales and the Slowdown in Bank Lending
by Rebecca Demsetz}
A complete understanding of the recent slowdown in bank lending requires an examination of loans that were originated by banks but are absent from their balance sheets as
a result of loan sales activity. Sales from domestic banks to foreign banks, nonbank financials, and nonfinancial institutions cause outstanding credit to leave the books of the
domestic banking system. If loan sales of this nature have increased (decreased) over
the "credit crunch" years, then trends in loans reported on the books of domestic banks
will overstate (understate) the severity of the recent credit slowdown.
This paper begins with an analysis of aggregate business loan sales by domestic
banks. We find that outstanding balances on loans sold to buyers outside the domestic
banking system increased in the late 1980s but decreased quite sharply between 1991
and 1992. We conclude that statistics describing outstanding business loans on the
books of domestic banks understate the severity of the recent slowdown in bank-originated business lending. Next, we take a closer look at the characteristics of loan sellers
and the conditions that foster loan sales activity. Since the liquidity provided by the secondary loan market may encourage loan origination, it is important to understand the
factors affecting secondary market activity. We use individual bank data to estimate a
model of loan sales activity. With these cross-sectional regressions, we identify several
bank characteristics that are statistically significant determinants of loan sales activity:
asset size, capital ratios, on-balance-sheet loans (as a fraction of assets), funding costs,
loan performance, loan portfolio concentration, and holding company affiliation. We
also explore the role of regional economic conditions in explaining cross-sectional variation in loan sales activity.2
1
2
The author thanks Akbar Akhtar, Richard Cantor, Beverly Mirlle. Rama Scth. John Simpson, and John
Wenninger for comments on earlier drafts of this paper. August Morct, Kathcrine Styponias. and Kevin
Lcyh provided excellent research assistance.
Both sections of this paper locus on the sale of business loans. Business lending is arguably the lending
area for which banks arc most likely to have superior origination and monitoring abilities. Analyses of
trends in business loan sales and the nature of the business loan sales market are thus of particular importance to our understanding of the slowdown in bank lending. Cantor and Demsetz (1993) report measures
131
Causes and Consequences
I. Accounting for Business Loan Sales When Evaluating the Slowdown in Bank Lending
The recent slowdown in bank lending to businesses is sometimes described by documenting a decline in the stock of commercial and industrial (C&I) loans on the books
of domestic banks (U.S. banks and U.S. branches and agencies of foreign banks).3 A
more correct evaluation of the decline in bank-originated business credit should account
for loans sold by domestic banks to foreign offices of foreign banks, nonbank financials,
and nonfinancial institutions since these loans do not appear on the books of selling
banks but do contribute to the stock of domestically originated bank credit. In addition,
it is useful to account separately for the sale of bank loans to nonfinancial institutions
since these sales represent a decrease in the outstanding bank credit of the nonfinancial
sector. This type of analysis is possible for the set of banks included in the Federal Reserve Board's Senior Loan Officer Survey of Lending Practices (SLP) because the SLP
asks banks to describe the types of institutions purchasing their loans as well as the
quantity of outstanding loans that have been sold. The SLP data do have two drawbacks, however: the sample of approximately sixty included banks is relatively small,
and not every bank reports loan sales data in each survey.
Reports of Income and Condition (Call Reports) facilitate a second analysis based
on data describing the full population of U.S. commercial banks;4 however, two factors
complicate the use of Call Report loan sales statistics. First, Call Reports do not ask
banks to identify the types of institutions buying their loans. Calculation of loan sales
from U.S. banks to institutions outside the U.S. banking system is thus less straightforward than in the case of the SLP data. Fortunately, Call Reports include statistics on
loan purchases as well as loan sales. By defining a sample that consists of all U.S.
banks, and then subtracting aggregate loan purchases from aggregate loans sales, it is
possible to use Call Report data to measure net business loan sales to institutions outside
the U.S. banking system. However, Call Reports do not facilitate the measurement of
loan sales to particular purchaser groups such as nonfinancial institutions.
A second problem associated with the Call Report sales data is that Call Reports
convey information on business loan sales and purchases in the form of flows rather
than stocks; that is, banks are asked to report the dollar volume of loans sold and loans
purchased in a given quarter. A flow-to-stock conversion of Call Report data is not
possible because of limited information regarding the duration of sold and purchased
loans. In particular, there is no way to determine the dollar volume of quarterly sales
or purchases still outstanding at the close of a given calendar quarter. Thus, while Call
Report data are a valuable source of information for analyzing trends in aggregate sales
activity, only the SLP provides data specifically suited for calculation of outstanding
balances on loans sold to institutions outside the domestic banking system, or to particular buyer groups such as nonfinancial institutions. The following section analyzes the
SLP (stock) data. Next we examine the Call Report (flow) data and summarize our
findings.
Footnote 2 continued
of credit growth that include mortgage, consumer, and business loans originated by intermediaries but
absent from their balance sheets because of direct loan sales or the issuance of asset-backed securities.
132
3
Throughout the remainder of this paper, we distinguish "U.S. banks" from "domestic banks." We use
"domestic banks" to denote U.S. banks and U.S. branches and agencies of foreign banks.
4
U.S. branches and agencies of foreign banks do not provide loan sales data in the Call Reports.
Analysis of the Survey of Lending Practices (SLP) Data
Since 1986, the SLP has asked responding banks to report approximately annually on
the outstanding amounts of C&I loans originated and then sold in their entirety or participated to others (without recourse). Bernanke and Lown (1991) compile these sales
statistics through June 1991. They observe a sharp peak and subsequent decline in this
loan sales series around the June 1990 survey. We extend the Bernanke/Lown analysis
using data from June 1992 and information on the types of institutions buying loans.
Loan sales involving a transfer from one U.S. bank to another do not remove loans from
the books of the U.S. banking system. Following Berger and Udell (1992), we use SLP
data to estimate trends in loan sales from U.S. banks to buyers other them U.S. banks.
The stacked bar chart in Chart 1 reports outstanding balances on commercial and industrial loans sold by SLP banks to buyers of several types: U.S. banks, U.S. branches
and agencies of foreign banks, foreign offices of foreign banks, nonfinancial institutions, and all other buyers (including finance companies, insurance companies, pension
funds, mutual funds, and bank trust departments). We refer to this final group of buyers
as "nonbank financials." Total outstanding balances on sold loans rise from $26 billion
in December 1985 to $80 billion by June 1990, but fall back down to $55 billion by June
1992. At each date, there is strong representation from each of three buyer groups: U.S.
banks, U.S. branches and agencies of foreign banks, and nonbank financials. The June
1992 SLP reveals an increase in the share of loan sales to U.S. banks and a sharp decrease in the share of loan sales to nonbank financials, which had been increasing from
June 1989 through June 1991.
Charts 2a, 2b, and 2c track outstanding balances on loans sold by SLP banks to three
particular buyer groups. Chart 2a reports loans sold to "non-U.S.-bank" buyers. These
buyers include nonfinancial institutions, nonbank financials, foreign offices of foreign
Chart 1: Outstanding Balances on Commercial and Industrial Loans Sold by LPS Banks
Billions of dollars, end of quarter values
Source: Senior Loan Officer Survey of Bank Lending Practices.
Note: A consistent buyer categorization is available beginning in 1988.
133
Causes and Consequences
Chart 2A: Outstanding Balances on Loans Sold by LPS Banks to Non-U.S Bank Buyers
Billions of dollars
60
Source: Senior Loan Officer Survey of Bank Lending Practices.
Note: "Non-U.S. Bank Buyers" includes nonfinancials, nonbank financials, foreign offices of
foreign banks, and U.S. branches and agencies of foreign banks.
Chart 2B: Outstanding Balances on Loans Sold by LPS Banks to Non-Domestic
Bank Buyers
Billions of dollars
60
Source: Senior Loan Officer Survey of Bank Lending Practices.
Note: "Non-Domestic Bank Buyers" includes nonfinancials, nonbank financials, and foreign
offices of foreign banks.
134
banks, and U.S. branches and agencies of foreign banks. Chart 2b reports loans sold to
"non-domestic-bank" buyers. These buyers include nonfinancial institutions, nonbank
financials, and foreign offices of foreign banks (but not U.S. branches and agencies of
foreign banks). Both series decline sharply between June 1991 and June 1992. Hence,
accounting for loan sales to non-U.S.-bank buyers or to non-domestic-bank buyers will
increase the measured decline in C&I lending by SLP banks, especially over the latter
"credit crunch" years. An important caveat regarding this conclusion is that there is
some inconsistency in the SLP response group. The report accompanying the 1991 SLP
explains that the decline in total loan sales from 1990 to 1991 "in part reflects the inability of some banks that had reported outstanding loans in the August 1990 survey to do
so |that| year." The same complication may bias the 1991-92 decline.5
Chart 2c tracks outstanding balances on loans sold by SLP banks to nonfinancial institutions only. These balances, measuring outstanding bank credit returned to the nonlinancial sector, are small compared with those depicted in Charts 2a and 2b, and thus
appear relatively constant over the time period examined.6
We estimate outstanding balances on loans originated and sold by all U.S. banks by
scaling the SLP loan sales data. Using Call Report data, we calculate the fraction of
quarterly commercial bank loan sales flows attributable to SLP banks, and use this scal5
Recorded annually, the total number ol SLP banks reporting sales volume reaches a minimum of 53 and a
maximum of 57 over the 1987-92 time period. This range may overstate the severity of reporting fluctuations, however, since the largest SLP banks tend to be the most active loan sellers as well as the most consistent reporters of SLP data. In 1991, 55 banks reported sales volume. In 1992. that number rose to 57.
6
A closer look at loans sold to nonfinancial institutions reveals a steady rise from March 1987 through June
1990, followed by a 50 percent decline between June 1990 and June 1991. However, by June 1992. loans
sold from SLP banks to nonlinancials had risen 22 percent above the value reported one year earlier.
Chart 2C: Outstanding Balances on Loans Sold by LPS Banks to Nonfinancial
Institutions
135
Causes and Consequences
ing ratio to inflate the SLP statistics reported in Charts 2a, 2b, and 2c. Though we believe that our approach is appropriate given the available data, there are three caveats
worth noting: first, our scaling factor is based on flow data rather than stock data; second, the same scaling factor is applied to all loan sales, regardless of purchaser identity;
and third, since the SLP responses of individual banks are confidential, our scaling factor cannot reflect changes in the set of SLP banks reporting sales volume.
The magnitude of our scaling factor is very high, averaging about 90 percent over
the relevant period; the SLP contains an overrepresentation of large banks, and these
banks are the most active participants in the loan sales market. The scaling factor is also
quite stable over time, ranging from 82 to 92 percent over the relevant period. Trends
reported in Charts 2a, 2b, and 2c are thus fairly representative of trends estimated for
the U.S. banking system as a whole.
Table 1 uses our estimates of outstanding balances on loans sold by all U.S. banks to
adjust trends in the on-balance-sheet lending of U.S. banks and the on-balance-sheet
lending of domestic banks. The first section of Table 1 adjusts the on-balance-sheet
lending of U.S. banks by adding outstanding balances on loans sold to all buyers other
than U.S. banks. The 1990-92 decline in lending by U.S. banks increases in magnitude
from 12.2 percent to 14.1 percent after accounting for relevant loan sales.
A similar pattern emerges from the second section of Table 1, where we adjust the
on-balance-sheet lending of U.S. banks and U.S. branches and agencies of foreign banks
and find that the 1990-92 decline in lending by these institutions increases in magnitude
from 6.0 percent to 7.6 percent after accounting for relevant loan sales. Since the onbalance-sheet statistics reported in the second section of Table 1 already reflect loan
sales from U.S. banks to U.S. branches and agencies of foreign banks, our loan sales adjustment includes only those sales made by U.S. banks to foreign offices of foreign
banks, nonbank financials, and nonfinancial institutions. A precise adjustment should
also reflect sales from U.S. branches and agencies to these buyers. We lack sufficient
data to calculate these sales for each year examined; however, data from the 1991 and
1992 SLPs suggest that sales by U.S. branches and agencies of foreign banks to foreign
offices, nonbank financials, and nonfinancial institutions are very small relative to U.S.
bank sales to these same institutions.
The third section of Table 1 adjusts C&I loans on the books of U.S. banks to reflect
only sales to nonfinancial institutions. In this case, the magnitude of the 1990-92 decline in C&I lending is practically unchanged, rising in magnitude from 12.2 percent to
12.4 percent. The stock of C&I credit returned to the nonfinancial sector through loan
sales is small compared with that transferred to other non-domestic-bank buyers.
Analysis of Call Report Data
Beginning in the second quarter of 1983, Call Reports provide individual bank flow data
describing loans originated by U.S. banks that have been sold or participated to others
(without recourse) during the calendar quarter ending with the report date.7 A second
series, describing loan purchases, begins in the second quarter of 1987. Since both series are constructed as flow variables rather than stock variables, they are not directly
comparable to the SLP sales series or to series describing the stock of outstanding business loans. Still, Call Reports provide a valuable source of information on aggregate
7
136
The term "loans originated" covers any loan made directly by the reporting bank. It does not include loans
purchased from others. In the case of syndicated loans, "loans originated" covers only the reporting bank's
share, even if the reporting bank is the lead bank in the syndication.
loan sales activity and can be used to confirm the pattern of declining business loan sales
outside the U.S. banking system over the recent credit slowdown. 8
Since banks do not describe the types of institutions purchasing the loans they have
sold, measurement of sales outside the U.S. banking system is less direct than in the case
8
The Call Report sales and purchase series exclude loans secured by onc-to-four-family residential properties and loans to individuals for household, family, and other personal expenditures. The reported series
thus describe business loans and some other loan types. However, the April 1986 description of the February 1986 SLP reports that "most bankers contacted by Board staff indicated that domestic commercial and
industrial was by far the most important category of loans sold." Consequently, we follow Bernanke and
Lown (1991) and Gorton and Haubrich (1990) in interpreting the Call Report series as indicators of business loan sales and purchase activities.
Table 1: Lending Growth, before and after Adjusting On-Balance-Sheet Lending for
Loan Sales
Outstanding Amounts,
Billions of Dollars
Percentage Change
1988-11
1990-11
1992-11
1988-11 to
1990-11
1990-11 to
1992-11
597.4
618.4
543.2
+3.5
-122
other than U.S. banksb
39.2
56.4
36.5
+43.9
-35.3
Total
636.6
674.8
579.7
+6.0
-14.1
Loans on the balance
sheets of U.S. banks
and U.S. branches
and agencies of foreign banksa
715.5
756.7
711.6
+5.8
-6.0
Loans sold to buyers
other than U.S. banks
and U.S. branches
and agencies of foreign banksb
20.9
27.0
12.5
+29.2
-53.7
Total
736.4
783.7
724.1
+6.4
-7.6
Loans on the balance
sheets of U.S. banksa
597.4
618.4
543.2
+3.5
-12.2
Loans sold to nonfinancial institutionsb
3.5
7.1
4.6
+103
-35.2
600.9
625.5
547.8
+4.1
-12.4
Parti
Loans on the balance
sheets of U.S. banks a
Loans sold to buyers
Part II
Part III
Total
a
- Source: Federal Reserve Bulletin.
b
- Estimates, based on Senior Loan Officer Survey of Lending Practices and Call Reports.
137
Causes and Consequences
of the SLP banks. In addition, Call Report data do not facilitate measurement of loan
sales to particular buyer groups such as nonfinancial institutions. However, by analyzing a sample consisting of all U.S. banks and subtracting aggregate business loan purchases from aggregate business loan sales, one can use the Call Report loan sales and
purchase series to measure trends in net business loan sales from U.S. banks to institutions outside the U.S. banking system.
Some examples help to illustrate how the Call Report loan sales and purchase series
can be used to measure the relevant trends. First consider a loan sold by a U.S. bank to
an institution outside the U.S. banking system. Since we analyze only those institutions
that are part of the U.S. banking system, this transaction contributes to aggregate sales
but not to aggregate purchases; the net sales figure will reflect this transaction. In contrast, consider a loan sold by one U.S. bank to another U.S. bank. This second transaction will contribute to both aggregate sales and aggregate purchases, and thus make no
contribution to the final net sales figure. Finally, a loan sold by an institution outside the
U.S. banking system and purchased by a U.S. bank will appear as a loan purchase only
and thus contribute negatively to aggregate net sales.
Series A in Chart 3 shows the quarterly sales flow by insured U.S. commercial banks
from 1986 through 1992. Loan sales grew rapidly during the late 1980s, reaching almost $300 billion per quarter in 1989. After the third quarter of 1989, loan sales slowed
sharply, falling to about $90 billion per quarter by 1993. Series B in Chart 3 shows the
quarterly purchase flow reported by insured U.S. commercial banks. This series is small
and stable. The net sales flow — the flow to buyers other than insured U.S. commercial
banks — is similar to the gross sales flow.
It is difficult to compare the Call Report sales trends to the SLP sales trends because
of differences in reporting frequencies. However, the basic trends conveyed through the
SLP (stock) and Call Report (flow) data appear consistent, with the Call Report series
Chart 3: Commercial and Industrial Loan Sales and Purchases by U.S. Banks
138
Billions of dollars, quarterly flows
300
declining slightly earlier than the SLP series. Note, however, that the relative magnitudes of the total sales series (Series A) and the net sales series (Series C) in Chart 3 suggest an interesting contrast between the SLP (stock) and Call Report (flow) data.
Responses of SLP banks indicate that over the relevant period, the stock of loans purchased by non-U.S.-bank buyers generally accounts for about two-thirds of the total
stock of sold loans. In contrast, the Call Report data indicate that nearly all of the flow
of business loan sales is purchased by institutions other than U.S. banks.
There are at least two possible explanations for this disparity. First, it is possible that
banks in the SLP sample are simply more likely than others to sell loans to U.S. banks.
However, further analysis reveals that the disparity between SLP (stock) data and Call
Report (flow) data persists even within the SLP sample. A second potential explanation
for the stock data/flow data disparity involves a difference in loan maturities by purchaser. In particular, if loans sold to U.S. banks are of much longer maturities than loans
sold to other purchasers, then the SLP (stock) data will show significant outstanding balances on loans sold to U.S. banks while Call Report (flow) data show little sales activity
to U.S. bank buyers.
II. The Determinants of Loan Sales Activity
In this section, we explore the characteristics of loan sellers and the conditions that foster secondary market activity. Our analysis is based on Call Report data, which include
information on the quarterly flow of loan sales associated with individual institutions as
well as relevant bank characteristics potentially related to sales activity. Since Call Reports do not identify the types of institutions purchasing loans from a given buyer, we
focus on total loan sales, regardless of buyer identity. Following Pavel and Phillis
(1987), we scale each bank's loan sales by asset size and use the resulting sales ratio as
the dependent variable in a cross-sectional regression analysis. We examine the stability
of the regression coefficients over time and reach some preliminary conclusions regarding factors underlying the loan sales decline during the "credit crunch" period. The independent variables in our reduced form model are described below. They are
motivated by several mutually consistent theories of loan sales activity.
Asset Size. Large banks are the biggest sellers of bank loans. Since a small number
of large banks account for a large fraction of all bank loan sales, our regressions include
a set of asset size dummy variables. We differentiate between twelve size groups, defined
in Table 2, and thus allow for nonlinearities at both ends of the asset size distribution.
Outstanding Loan Volume. We explore the ratio of outstanding loans-to-assets in order to determine whether banks that hold a high fraction of assets as loans are more or
less likely to sell the loans that they originate.9 Haubrich and Thomson (1993) find a
significant positive relationship between the loans-to-assets ratio and the concurrent
sales-to-assets ratio. We include two measures of outstanding loan volume, the loansto-assets ratio and the C&I loans-to-assets ratio. The C&I loans-to-assets ratio is included because the Call Report loan sales data primarily reflect sales of C&I loans. We lag
both loan-to-assets ratios by one quarter, since their current values will reflect concurrent sales activity.
Funding Costs. To the extent that some banks have a comparative advantage in funding credit and others in originating credit, we should see a positive relationship between
9
Sec Mcster (1992) for a discussion of diseconomies of scope between traditional activities of originating
and monitoring loans and nonlraditional activities of loan selling and buying.
139
Causes and Consequences
140
funding costs and sales activity. In other words, banks with costly funding sources
should be more likely than others to sell the loans they originate, all else equal (see Pennacchi 1988). Following Haubrich and Thomson (1993) and Berger and Udell (1992),
we measure funding costs as the sum of brokered deposits and large deposits (deposits
that exceed the $100,000 FDIC insurance limit), scaled by assets. Again, a one-quarter
lagged value is used.
Bank Capital. Banks may sell loans in order to capture the benefits of loan origination without the costs of increased capital. In particular, capital-constrained banks may
use the loan sales market to facilitate loan origination, receiving origination and servicing fees without increasing the size of the loan portfolio they hold.10 We test the hypothesis of an inverse relationship between bank capital and loan sales activity by including
in our regression the ratio of equity capital to assets. Again, we use the one-quarter
lagged value because the current capital-to-assets ratio will reflect concurrent capitalenhancing sales activity. We also explore the effects of some additional capital measures designed to test more directly the hypothesis that capital-constrained institutions
are more likely than others to be active loan sellers. These additional capital measures
are defined later in this section.
Loan Portfolio Diversification. Another theory of loan sales suggests that banks sell
and buy loans to enhance the diversification of their loan portfolios (Pavel and Phillis
1987). The diversification motive implies that sales activity should be most prevalent
among banks with loan origination focused in a narrow geographical area or business
line, all else equal. Unfortunately, it is difficult to distinguish these banks from other
banks. We lack information on the types of loans originated by specific institutions, and
information on the types of loans within existing portfolios is not detailed enough to
build fully satisfactory measures of loan concentration. In particular, it is not possible
to identify industrial and geographical concentrations, the two types of loan concentrations that are probably most relevant to the sale of business loans. Instead, we measure
overall portfolio concentration using a concentration index equal to the sum of the
squared shares of the loan portfolio in each of five Call Report loan categories: real estate, agriculture, consumer, C&I, and other loans. Again, we use a one-quarter lag.
Loan Portfolio Performance. Buyers' perceptions of poor loan quality may lead to
reduced demand for a particular bank's loans. Hence, the selling institution's past loan
portfolio performance may be positively related to the loan sales ratio. An alternative
supply-side hypothesis is that a bank with poor loan performance will try to sell loans
in order to maintain an existing level of capital. This hypothesis suggests a negative correlation between loan portfolio performance and sales activity. Our reduced form model
includes two measures of loan portfolio performance. The first measure, the nonperforming loan ratio, is defined as the sum of nonaccruing loans and loans past due ninety
or more days, divided by total loans. Again, a one-quarter lagged value is used. Both
Pavel and Phillis (1987) and Haubrich and Thomson (1993) examine a second measure
of portfolio performance, the ratio of net charge-offs to assets. Both studies find a positive relationship between net charge-offs and loan sales, though that relationship is statistically significant only in the former analysis. Along with the lagged nonperforming
loan ratio, we include the net charge-off ratio, equal to net charge-offs occurring in the
current quarter, scaled by assets.
Holding Company Affiliation. Haubrich and Thomson (1993) find that holding company affiliation is a positive, significant determinant of sales activity and explain that
10
See Pavel and Phillis (1987) for an early discussion.
loan sales are frequently transactions between sellers and buyers in the same holding
company. We include a dummy variable indicating whether a given bank belongs to a
holding company.
Economic Conditions. The role of economic conditions has been given little attention in previous studies of loan sales activity. However, there are at least two reasons to
believe that economic conditions, and especially expectations of future economic conditions, will affect loan sales activity. First, optimism regarding the economy should
stimulate loan origination, and strong loan origination creates the potential for strong
loan sales activity. Second, since economic strength should positively affect the performance of newly originated loans, optimism regarding the economy may dampen buyers'
fears of weak loan performance. Both stories suggest that a promising economic environment will stimulate loan sales activity.
To explore the importance of economic conditions in the context of our cross-sectional analysis, we construct measures of regional economic strength and test the hypothesis that at a given point in time, banks operating in promising economic
environments are more active loan sellers than those operating in weak economic environments, all else equal. We measure the strength of each regional economy using a
state-specific consumer confidence index, which we aggregate to the regional level.11
In summary, our regression equation expresses the loan sales-to-assets ratio as a
function of total assets, the loans-to-assets ratio, the C&I loan ratio, a measure of funding costs, the equity capital-to-assets ratio, a measure of loan portfolio diversification,
two measures of portfolio performance, holding company affiliation, and a measure of
regional economic conditions. Most of these variables are potential determinants of
the supply of sold loans; others are related to the demand for loans in the secondary
market.
Model Estimation
The model is estimated separately for the first quarter of each year between 1986 and
1992. The sample analyzed includes all insured domestic commercial banks. It would
be possible to pool the seven yearly data sets and estimate a single set of coefficients
using a panel model. An advantage of separate yearly regressions is that they allow us
to examine changes in estimated coefficients over time. Those changes prove to be important, indicating that the implicit assumption of coefficient stability underlying a panel model is not valid in this case.
We estimate the model using a tobit procedure. A tobit model is used when the regression's dependent variable combines both qualitative and quantitative information.
In this case, the qualitative information conveyed by the dependent variable is each
bank's participation or lack of participation in the loan sales market. The quantitative
information conveyed is the magnitude of sales activity (relative to total assets) for those
institutions that do participate.
1
' In particular, we create regional measures of economic strength by averaging state-specific values of the
Sindlinger "Household Money Supply Index," which reflects the fraction of surveyed households reporting that (1) total combined annual income is up from, or the same as, six months prior, (2) total combined
annual household income in the next six months will remain the same or be up, (3) the number of jobs
where the respondent works will increase or remain the same in the next six months, and (4) business conditions where the respondent lives will be the same or better in the next six months. Each region represents one Federal Reserve District. A state belonging to two Federal Reserve districts contributes to the
value calculated for each district. Since data were unavailable for Alaska and Hawaii, the District 12 values do not reflect these two states.
141
Causes and Consequences
Results and Discussion
142
Table 2 reports our results. In one sense the results are encouraging, since most independent variables are highly significant in each year examined and most of the estimated
coefficients have the hypothesized signs. However, some of the estimated coefficients
vary over time, making it difficult to draw firm conclusions regarding certain determinants of loan sales activity.
As expected, asset size is an important determinant of sales activity. Since the size
category deleted from the regression corresponds to banks of intermediate size (those
with assets greater than $ 100 million but less than $500 million), the significant positive
coefficients on both the highest and lowest size dummies suggest an interesting nonlinear relationship: sales activity increases at the high end of the size distribution and at the
low end of the size distribution, all else equal, though the increase is of much greater
magnitude at the high end. This pattern is consistent with the need for small banks to
participate in the secondary market in order to meet lending limits or to enhance portfolio diversification. Although the often negative coefficient associated with the concentration index would suggest that more diversified banks are less likely to be active
sellers, that index is a rather crude measure of the type of portfolio concentration that is
likely to stimulate sales activity, as noted earlier. Indeed, its coefficient is quite unstable
across the yearly regressions reported in Table 2.
A closer look at the asset size coefficients reveals that their relative magnitudes
change substantially across the seven-year period examined. In particular, coefficients
associated with the largest size groups increase in the middle of the seven-year period,
peaking in 1989, concurrently with aggregate loan sales. This pattern suggests that the
regression model excludes some important factor(s) that differentially affect(s) the sales
activity of small and large banks. A likely candidate is the origination of large mergerrelated loans, loans that are frequently divided into smaller credits and sold in the secondary market.12 Unfortunately, we lack historical bank-level data on merger-related
originations, so the regression model cannot control for this factor.13
Outstanding loan volume is significantly related to sales activity, with positive coefficients associated with each of the loans-to-assets variables in each year examined.
Also consistent across the seven regression equations are the positive, significant coefficients associated with the variables measuring funding costs and holding company affiliation. The relationship between the nonperforming loan ratio and sales activity is
generally negative and significant; the coefficient associated with the net charge-off ratio
is less robust. The capital ratio coefficient is negative and highly significant (as hypothesized) in each year. Finally, our measure of regional economic conditions enters the
equation with a positive and highly significant coefficient in 1988-92 but a negative significant coefficient in the first two years examined, suggesting that the link between economic conditions and sales activity may be more complex than initially hypothesized.14
Berger and Udell (1992) note that two large banks alone account for a sizable share
of the loan sales market in the 1986-92 period. When these banks are deleted from the
model, the results are qualitatively unchanged, although the asset size coefficients asso12
Indeed, a strong correlation between corporate merger activity and aggregate sales activity has been noted
(sec Berger and Udell 1992).
13
See Dcmsetz (1994) for a more detailed discussion.
14
Relationships between regional economic conditions, lending opportunities, and loan sales activity arc
explored further in Dcmsetz (1994).
Table 2: Yearly Tobit Models: Dependent Variable=(Sales/Assets)
Independent
Variables
(1)
1992-1
Coeff.
(t-stat)
(2)
1991-1
Coeff.
(t-stat)
(3)
1990-1
Coeff.
(t-stat)
(4)
1989-1
Coeff.
(t-stat)
(5)
1988-1
Coeff.
(t-stat)
(6)
1987-1
Coeff.
(t-stat)
(7)
1986-1
Coeff.
(t-stat)
Regional economic
conditions
.0021**
(11.8)
.0021**
(12.9)
.0012**
(11.2)
.0009**
(3.04)
.0007**
(4.36)
-.0011**
(3.74)
-.0003**
(4.90)
Assets < $10 m.
.0437**
(5.94)
.0343**
(9.03)
.0189**
(8.19)
.0377**
(8.28)
.0194**
(5.68)
.0142*
(2.08)
.0230**
(8.79)
$10 m. < assets < $25 m.
.0072
(1.67)
.0143**
(6.08)
.0079**
(5.26)
.0099**
(3.22)
.0090**
(3.83)
-.0011
(.234)
.0103**
(5.51)
$25 m. < assets < $50 m.
-.0030
(.759)
.0075**
(3.45)
.0042**
(2.98)
.0041
(1.41)
.0035
(1.54)
-.0135**
(2.92)
.0020
(1.12)
$50 m. < assets < $100 m.
-.0038
(.970)
.0016
(.703)
-.0005
(.341)
-.0029
(.994)
-.0018
(.766)
-.0134**
(2.81)
-.0017
(.933)
$500 m. < assets < $1 b.
.0163*
(1.96)
.0217**
(4.60)
.0058
(1.81)
.0116
(1.71)
-7.5e-07
(0.00)
.0043
(.372)
.0037
(.824)
$1 b. < assets $5 b.
.0249**
(2.98)
.0124**
(2.66)
.0057*
(1.90)
.0017
(.277)
.0080
(1.71)
.0232*
(2.30)
.0155**
(4.04)
$5b. < assets < $10 b.
.0385*
(2.51)
.0268**
(3.17)
.0191**
(3.42)
.0223
(1.95)
.0241*
(2.46)
.0526**
(2.61)
.0206**
(2.57)
$10 b. < assets < $20 b.
.0652**
(3.23)
.0474**
(3-94)
.0243**
(2.83)
.0268
(1.49)
.0305*
(2.07)
.0788*
(2.20)
.0200
(1.32)
$20 b. < assets < $50 b.
.0841*
(2.34)
.1098**
(5.24)
.1487**
(12.4)
.1564**
(5.58)
.2817**
(12.8)
.1754**
(3.85)
.0138
(.712)
Assets > $50 b.
.1770**
(4.76)
.1859**
(8.38)
.3114**
(19.8)
.4910**
(16.5)
.2271**
(9.09)
.1818**
(3.52)
.0707**
(3.64)
Loans/assets
.1079**
(10.1)
.0464**
(7.86)
.0438**
(11.5)
.0616**
(8.10)
.0520**
(9.09)
.1210**
(9.92)
.0653**
(13.5)
C&l loans/assets
.1225**
(6.13)
.0794**
(7.63)
.0697**
(11.3)
.1318**
(10.6)
.1077**
(11.7)
.1943**
(10.5)
.0806**
(11.6)
Capital/assets
-.2583**
(5.87)
-.1901**
(7.77)
-.0985**
(7.11)
-.2123**
(6.98)
-.1162**
(5.30)
.4191**
(10.5)
-.1058**
(5.20)
Nonperforming loans/
assets
-.3291**
(3.07)
-.2892**
(4.86)
-.2047**
(5.55)
-.3229**
(4.79)
-.1175**
(2.74)
.0186
(.235)
-.1578**
(5.16)
Net charge-offs/assets
-1.546*
(2.37)
.6153
(1.86)
-.4855**
(2.78)
-.5121
(1.51)
.1311
(.731)
-.3348
(.900)
.1234
(.921)
Concentration index
-.0306**
(2.58)
-.0202**
(2.97)
-.0024
(.541)
-.0294**
(3.19)
.0111
(1.57)
-.0476**
(3.10)
.0242**
(4.12)
Notes: Deleted size category is $100 m. < assets < $500 m.
* Significant at the 5 percent level.
** Significant at the 1 percent level.
(Continued)
143
Causes and Consequences
Table 2: Yearly Tobit Models: Dependent Variable=(Sales/Assets) (Continued)
144
(D
Independent
Variables
1992-1
Coeff.
(t-stat)
(2)
1991-1
Coeff.
(t-stat)
(3)
1990-1
Coeff.
(t-stat)
(4)
1989-1
Coeff.
(t-stat)
(5)
1988-1
Coeff.
(t-stat)
(6)
1987-1
Coeff.
(t-stat)
(7)
1986-1
Coeff.
(t-stat)
(Large + brokered dep)/
assets
.0341"
(2.69)
.0366**
(5.38)
.0296**
(8.71)
.0416**
(5.01)
.0264**
(4.15)
.0598**
(5.91)
.0564**
(11.7)
Belongs to holding
company
.0255"
(8.02)
.0161**
(9.23)
.0101**
(9.19)
.0248**
(11.1)
.0210**
(12.4)
.0449**
(13.2)
.0172**
(13.4)
Constant
-.2068**
(18.1)
-.1401**
(19.1)
-.1142**
(17.9)
-.1464**
(8.08)
-.1435**
(11.6)
-.1744**
(10.1)
-.0820**
(16.1)
11,685
12,089
12,428
12,783
13,305
13,794
12,969
757
999
1,515
1,106
1,155
934
1,654
Number of observations
Chi-squared
Notes: Deleted size category is $100 m. < assets < $500 m.
* Siignificant at the 5 percent level.
** Significant at the 1 percent level.
ciated with the "assets > $ 50 billion" size group are, not surprisingly, much smaller than
in Table 2, especially in 1989, at the height of aggregate loan sales activity.
We also experimented with some alternative capital ratios. For 1991 and 1992, we
constructed the following variables: tier 1 capital divided by assets (leverage ratio); tier
I capital divided by risk-weighted assets (tier 1 ratio); tier 1 plus tier 2 capital divided
by risk-weighted assets (total capital ratio); and a dummy variable (capital-constrained) which equals one for banks with a total capital ratio less than 8 percent, or tier
1 or leverage ratios less than 4 percent, and equals zero for all other banks.15 According
to the capital constraint hypothesis discussed above, coefficients associated with these
capital ratios should all be negative, while the coefficient associated with the "capitalconstrained" dummy should be positive. When data from 1992 are used, hypotheses regarding the capital ratios are confirmed, although the coefficient associated with the
"capital-constrained" dummy is not significantly different from zero. A similar pattern
arises with 1991 data, although in this case, the "capital-constrained" dummy variable
is significant at the 10 percent level.
Cross-sectional results reported in Table 2 provide some insight into the decline in
loan sales over the "credit crunch" years. Our regression results indicate that even after
we control for bank-specific characteristics, many of which reflect the economic environment in which the bank operates, strong regional economies are associated with
strong loan sales activity in all but the first two years examined. To the extent that these
cross-sectional findings are generalizable to changes across time, we should expect loan
sales activity to decline in a recessionary period. Indeed, estimates from the 1990 regression in Table 2 suggest that changes in economic conditions on the order of those
that occurred over the recent economic downturn can have a substantial effect on the
'• Banks that fail to meet any of these three conditions are considered "undercapitalized" under the "Prompt
Corrective Action" provisions of the Federal Deposit Insurance Corporation Improvement Act of 1991.
dollar volume of loan sales. Replacing 1990 values of the consumer confidence index
with 1991 values causes aggregate fitted sales to fall by over 10 percent, based on the
1990 regression coefficients.16 This change is much larger in magnitude than those that
result when 1990 values of the other regression variables are replaced by their 1991 values. These results are summarized in Table 3.
III. Summary
Aggregate data reveal a strong decreasing trend in loan sales from U.S. banks to nonU.S.-bank buyers and from domestic banks (U.S. banks and U.S. branches and agencies) to non-domestic-bank buyers over the "credit crunch" years. One important implication of this decreasing trend is that statistics describing outstanding business loans
on the books of U.S. banks and on the books of all domestic banks actually understate
l6
Fitted values of the salcs-to-assets ratios are calculated as described in Maddala (1983), pp. 159-60, equation 6.37. Fitted aggregate sales are then calculated by multiplying each bank's fitted sales-to-asscts ratio
by bank size and summing across all observations.
Table 3: Sensitivity of Aggregate Fitted Sales to Changes in the Values of Regression
Variables
Each row reports the percentage change in fitted aggregate sales when 1990-1 values for one (or some) of the
regressors are replaced by 1991-1 values and all other variables are held constant at their 1990-1 levels. Results
are derived using coefficients from the 1990-1 regression in Table 2. Fitted values of the sales-to-assets ratios are
calculated as described in Maddala (1983), pp. 159-60, equation 6.37. Fitted aggregate sales are then calculated
by multiplying each bank's fitted sales-to-assets ratio by bank size and summing across all observations.
Independent Variable(s)
Percentage Change in Fitted Aggregate Sales,
Replacing 1990-1 Regressor Values by
1991-1 Values
Regional economic conditions
(regional consumer confidence)
-10.2
Asset size
(dummy variable indicating asset size group)
+0.9
Outstanding loan volume
(loans/assets, C&l loans/assets)
+0.1
Funding costs
[(large + brokered deposits)/assets]
+0.2
Bank capital
(capital/assets)
-0.2
Loan portfolio concentration
(concentration index)
Loan portfolio performance
(nonperforming loans/assets, net chargeoffs/assets)
Holding company affiliation
(dummy variable: "belongs to holding
company")
0
-0.8
0
145
Causes and Consequences
the severity of the recent slowdown in bank lending. We estimate that the magnitude of
the 1990-11 to 1992-II decline in C&I lending by U.S. banks increases from 12.2 percent
to 14.1 percent after accounting for loans sold to non-U.S.-bank buyers and that the
magnitude of the decline in C&I lending by domestic banks (U.S. banks and U.S.
branches and agencies of foreign banks) increases from 6.0 percent to 7.6 percent after
accounting for loans sold to non-domestic-bank buyers. The adjustment for loans sold
to nonfinancial institutions is much smaller. Estimates based on the sales activities of
the SLP banks suggest that the stock of C&I credit returned to the nonfinancial sector
through loan sales is small compared with that transferred to other loan buyers.
Microdata describing loan sales by individual banks facilitate analysis of several determinants of loan sales activity. Cross-sectional regressions based on data from 198692 indicate that asset size, loans-to-assets ratios, capital ratios, funding costs, and holding company affiliation are consistently significant determinants of sales activity. The
effects of portfolio diversification and portfolio performance are less robust. A measure
of regional economic conditions is positively and significantly related to loan sales in
most of the years examined, and regression results based on 1990 data suggest that
changes in economic conditions on the order of those that occurred over the recent economic downturn may have a substantial effect on the dollar volume of loan sales.
References
146
Berger, Allen, and Gregory Udell. 1993. "Securitization, Risk, and the Liquidity Problem in Banking." In Michael Klausner and Lawrence White, eds., Structural Change
in Banking, pp. 227'-91.
Bernanke, Ben S., and Cara S. Lown. 1991. "The Credit Crunch." Brookings Papers
on Economic Activity, no. 2, pp. 205-47.
Cantor, Richard, and Rebecca Demsetz. 1993. "Securitization, Loan Sales, and the
Credit Slowdown." Federal Reserve Bank of New York Quarterly Review, vol. 18,
no. 2, pp. 27-38.
Demsetz, Rebecca S. 1994. "Economic Conditions, Lending Opportunities, and Loan
Sales." Working Paper no. 9403, Federal Reserve Bank of New York, February 1994.
Gorton, Gary B., and Joseph G. Haubrich. 1990. "The Loan Sales Market." In G. Kaufman, ed., Research in Financial Services: Private and Public Policy, vol. 2, pp. 85-135.
Haubrich, Joseph G., and James B. Thomson. 1993. "Loan Sales, Implicit Contracts,
and Bank Structure." Working Paper, Federal Reserve Bank of Cleveland, July
1993.
Maddala, G.S. 1983. Limited-dependent and Qualitative Variables in Econometrics.
Cambridge: Cambridge University Press.
Mester, Loretta J. 1992. "Traditional and Nontraditional Banking: An InformationTheoretic Approach." Journal of Banking and Finance, vol. 16, pp.545-66.
Pavel, Christine, and David Phillis. 1987. "Why Commercial Banks Sell Loans: An
Empirical Analysis." Economic Perspectives, Federal Reserve Bank of Chicago,
May/June, pp.3-14.
Pennacchi, George. 1988. "Loan Sales and the Cost of Bank Capital." Journal of Finance, vol. 43, no. 2, pp. 375-96.
147
Causes and Consequences
148
Foreign Credit Expansion in the United States
by Rama Seth1
Several recent studies have examined U.S. bank credit growth during the recent recession.2 These studies have focused on the balance sheet contraction of U.S. banks and
the reasons for this contraction. The role of foreign banks during this period, however,
has largely been ignored. This article seeks to assess the prevalence of foreign bank
lending in the United States in 1990-92 and to analyze the determinants of that lending.
We consider the behavior of foreign banks as a group, then identify differences in foreign bank behavior across countries and institutional forms. We give particular attention to the factors that led certain groups of banks to increase credit.
On the surface, it appears that foreign bank lending in the United States increased
while domestic banks were reducing loan extensions.3 Although foreign banks taken
together did increase their U.S. lending, differences emerge when the group is disaggregated according to both institutional form and nationality. While the U.S. branches and
agencies of foreign banks expanded their outstanding commercial loans, foreign bank
subsidiaries followed the pattern of U.S.-owned banks and contracted their loans. Early
in the recession, Japanese banks diverged from domestic banks and many other foreign
banks by increasing their U.S. lending; later, French and Canadian banks would do the
same.
In Section I, we first briefly discuss our sample selection. The second section describes the growth in balance sheets of foreign banks in the aggregate and then of two
broad groupings of these banks—namely, growing and retrenching foreign banks. Section III discusses various attributes of these bank categories—capital strength, asset
quality, output growth in the home country, and relative lending rates in the home coun1
1 am grateful to Akbar Akhtar, Ron Johnson, Robert McCaulcy, and Anthony Rodrigues for comments and
suggestions, to Ted Fischer for research assistance, and to Valerie LaPorte for editorial assistance.
2
For a summary of these studies, see Allen N. Berger and Gregory F. Udell, "Did Risk-Based Capital Allocate Bank Credit and Cause a 'Credit Crunch' in the U.S.?" Journal of Money, Credit, and Banking, vol.
26, 1994.
3
The growth in credit by foreign banks during this period can only partially be attributed to loan purchases
from domestic banks. At least some of the growth must be attributed to new lending.
149
Causes and Consequences
try—and examines the differences in these attributes across bank categories. We tentatively assess some determinants of foreign lending by examining means differences in
this section. In section IV, we present regression models that not only test the capital
adequacy hypothesis directly but also measure the extent of lending differences by nationality. Section V examines the factors that may explain nationality differences and
specifies a reduced-form model. Section VI discusses the results of this model, distinguishing between the factors explaining growth in branches and agencies, on the one
hand, and subsidiaries of foreign banks, on the other. The paper's final section lays out
the conclusions of the study.
We find that the dissimilarity in lending behavior between foreign and domestic
banks also extends to a dissimilarity in the determinants of this lending. Two supplyside factors that have been found to be significant in explaining recent U.S. bank credit
behavior are capital strength and asset quality. While capital strength was found to be
significant in explaining lending by the positive growth group of branches and agencies,
it did not appear to be important in explaining subsidiaries' lending. Reported asset
quality did not emerge as a significant determinant of lending by any group of foreign
banks. Instead, a desire to increase market share, better U.S. lending rates, and credit
diversion from offshore to onshore branches appear to explain the growth in foreign
bank branches and agencies. In the case of the few subsidiaries that increased credit,
market share is one plausible explanation underlying their behavior.
I. Sample Selection
The sample was dictated by data availability. In each instance, we tried to use all the
information available for that particular analysis; consequently, the samples varied
somewhat across analyses. For data analyses without capital ratio variables, all foreign
banks were used.4 For analyses using the capital ratio, data limitations required the sample to be more limited.3 In the tables for these analyses, subsidiaries were included only
if they reported capital ratios at the end of December 1990. Since the reporting of capital ratios by parents of foreign banks was somewhat more sparse, we included the
branches and agencies provided that their parent had reported capital ratios at least once
during 1990-91. Because certain variables used in the regressions were only available
for selected industrial countries, the final table with detailed regression results only included foreign banks from Belgium, Canada, Denmark, Finland, France, Germany, Italy, Japan, Netherlands, Norway, Spain, Sweden, Switzerland, and the United Kingdom.
Since the aim of this paper was to examine credit behavior of foreign banks over the
course of the recession, in most of our analysis the time period used spanned the recession. On the assumption that some foreign banks may have needed time to adjust to recessionary conditions, we generally conducted the analyses through the date of the most
recent data availability (at the time the data work was done—that is, September 1992).
II. Credit Growth
150
Foreign banks as a group continued to increase corporate loan extensions in 1990-92
when domestically owned banks were contracting theirs (Chart 1). Differences in lending behavior across foreign banks appear, however, when these banks are subdivided by
institutional form.
4
Sec Table 1; the first, third, and fourth columns of Table 2; Tables 3 and 3a; and appendix Tables 1 and 2.
5
Including the second (capital ratio) column of Table 2, Tables 4, R-l, and R-2.
Foreign banks provide services to U.S. customers through seven different institutional forms.6 The two most important forms—branches and agencies of foreign banks and
foreign bank subsidiaries—behaved rather differently during the recent bank credit
slowdown. 7 As a group, U.S. branches and agencies of foreign banks expanded their
outstanding loans. By contrast, the U.S.-chartered foreign banks, or subsidiaries, followed the same course as U.S.-owned banks in reducing their lending. Indeed the reduction in lending by these foreign banks (which account for less than a third of credit
extended to U.S. corporations by all foreign banks in the United States) from the start
of the recession to September 1992 was almost identical to the reduction in lending by
U.S.-owned banks (Table 1). Because these two groups of banks are alike in their portfolio composition, clientele, and other characteristics, the similarity in their lending behavior during this period is not surprising.8
Even after institutional differences are taken into account, all foreign banks did not
behave homogeneously. The difference in behavior is most easily observed by dividing
banks into two categories, those whose home country banks as a group expanded their
total U.S. corporate loan portfolio after the start of our recession and those whose home
country banks as a group decreased their portfolio. We shall term these categories "positive growth" banks and "negative growth" banks, respectively.
6
For a summary of the different institutional types, their functions and their relative importance, see Rama
Scth, "Foreign Banks1 Contribution to Excess Capacity," FRBNY memo, November 1992.
7
Ronald Johnson notes this difference in 'The Bank Credit 'Crumble,'" Federal Reserve Bank of New York
Quarterly Review, Summer 1991.
8
Sec Scth and Quijano, "Japanese Banks' Customers in the United States," Federal Reserve Bank of New
York Quarterly Review, Spring 1991; and Scth, "Excess Capacity."
Chart 1: Commercial and Industrial Loan Growth by U.S. and Foreign-Owned Banks
June 30, 1990 = 100
120
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Notes: Tick marks refer to the end of the quarter. The numbers in parentheses are the values of commercial and industrial loans as of the end of 1992. All foreign banks include U.S.
branches and agencies and subsidiaries of foreign banks.
151
Causes and Consequences
Table 1: Lending of Banks by Institutional Type and Pattern of Growth
Growth of Commercial and Industrial Loans from June 30,1990, to September 30,1992 (Percent)
Domestically Owned
Domestic Banks
Subsidiaries of
Foreign Banks
Branches and Agencies
of Foreign Banks
Total
-3.03
-11.58
15.78
Positive growth
31.88
23.73
126.04
Negative growth
-25.66
-14.10
-7.21
Breakdown by Commercial and Industrial Loans as of September 30,1992
Domestically Owned
Domestic Banks
Subsidiaries of
Foreign Banks
Branches and Agencies
of Foreign Banks
400.7
41.6
132.4
52.86
9.33
33.68
47.14
90.67
66.32
Domestically Owned
Domestic Banks
Subsidiaries of
Foreign Banks
Branches and Agencies
of Foreign Banks
11,201
139
522
50.17
13.67
53.83
49.83
86.33
46.17
Total
(billions of dollars)
Positive growth
(percent of above total)
Negative growth
(percent of above total)
Breakdown by Number of Banks as of September 30,1992
152
Total
(number of banks)
Positive growth
(percent of above total)
Negative growth
(percent of above total)
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Notes: The definition of positive and negative growth bank depends on the type of bank. For
subsidiaries and branches and agencies, all the banks were grouped by the country of their
owners. Then each bank was classified by whether the banks from the home country as a
whole increased or decreased commercial and industrial loans to the United States from June
30,1990, to September 30,1992. For domestically owned banks, the classification was done
by the growth in each individual bank's loans over the same period. This means that the statistics of domestically owned banks are those in existence over 1990-92 (a constant sample),
while this is not necessarily true for foreign-owned banks. The total figures for domestically
owned banks exclude those not in the constant sample. The total figures for foreign banks
exclude banks from countries that did not maintain a continual presence in the U.S. from 199092. The positive growth subsidiaries exclude the expansion in holdings by a Spanish bank in
First Fidelity Inc. in the first quarter of 1991. This expansion caused the banks controlled by
First Fidelity to be counted as subsidiaries. The positive growth branches and agencies exclude
the New York branch of the Union Bank of Switzerland in the New York and Atlanta agencies of
the Bank of Nova Scotia. These entities had disproportionately large increases in loans.
The resemblance of subsidiaries to domestic banks remains when banks are grouped
according to this scheme. The increase in lending by branches and agencies of foreign
banks that grew far exceeds that of growing domestic banks or the small group of growing subsidiaries. By the same token, the branches and agencies that contracted reduced
their lending less than a similar group of subsidiaries and domestic banks (Table 1 ). 9
Indeed, one could argue that subsidiaries conform more closely to the stereotype of
domestic banks than do domestic banks themselves. A far smaller percentage of subsidiaries, both by number of banks and share of loans, showed positive growth than either the branches and agencies or the domestic banks. Growing subsidiaries account for
less than 10 percent of all foreign bank subsidiaries, by value of commercial and industrial loans or by number of banks.
Another revelation of this grouping10 is that a higher percentage of U.S.-owned
banks than of branches and agencies increased their lending if the value of such lending
is used as the measure. Roughly half of the U.S.-owned banks, by value of loans, fell
into the positive growth group, while only a third of foreign bank branches and agencies
fell into this group (Table 1 and Chart 2). Even if we calculate the percentage according
to the number of banks, branches and agencies that grew did not much exceed domestic
banks that grew. By contrast, as pointed out earlier, the percentage of subsidiaries that
grew was very small.
In each of the two growth categories that we have defined, individual countries have
had wide-ranging growth rates in the corporate credit extended by their banks. Germany, Australia, Canada, and France are among the countries that fall into the positive
growth group, while Japan, the United Kingdom, and Italy fall into the negative growth
group. An appendix table details the growth in corporate loans by banks from each
country during several periods.
Foreign bank commitment to the U.S. market appears much larger if we add the assetbacked commercial paper sponsored by foreign banks to their direct lending to U.S. corporations. This type of "lending" is capital economizing and is engaged in most by the
least capitalized U.S. banks. If we factor in the commercial paper sponsored by Japanese
banks, one group of banks for which we have the data, the results are indeed striking. Although six of ten of these banks decreased their commercial and industrial loans, their
total commitment to the U.S. market can be said to have increased if we include the assetbacked commercial paper that they sponsored (Chart 3). Data limitations, however, force
us to exclude this aspect of foreign bank "lending" in the analysis that follows.
Foreign banks also lend to U.S. corporations by booking loans outside the United
States. A previous study concluded that these loans formed a substantial part of the total
growth in lending by foreign banks (see Chart 4). 11 More recently, however, such lending has not grown as quickly because the need to circumvent reserve requirements no
longer motivates banks to book loans through offshore centers. Again because of data
limitations, we will not include these loans in the analysis below.
9
The figures in Table 1 exclude the subsidiaries that were formed as a result of takeovers by a Spanish
acquirer. These banks were all part of the same holding company— First Fidelity— and their inclusion
tends to skew our results a great deal. When these subsidiaries are included, lending growth by positive
growth subsidiaries increases to a phenomenal 136 percent over the period. As a consequence, subsidiaries appear to resemble branches and agencies rather than the domestic banks. But because so few subsidiaries show positive growth, we chose to highlight their similarity to domestic banks in this instance also.
10
It is commonly believed that foreign banks continued to lend as domestic banks cut back during the credit
slowdown in the United States.
1
* See Robert N. McCauley and Rama Seth, "Foreign Bank Credit to U.S. Corporations: The Implications of
Offshore Loans," Federal Reserve Bank of New York Quarterly Review, vol. 17, no. 1 (Spring 1992).
153
Causes and Consequences
III. Mean Attributes
Studies of domestic banks have shown a connection between particular attributes of
banks and banks' lending behavior. If applied to foreign bank lending in the United
States, the findings of these studies would suggest that, ceteris paribus, banks from
countries that increased credit would likely have better asset quality and a stronger capital base than banks from countries that retrenched credit in the United States. In addition, demand considerations would suggest that positive (U.S. credit) growth groups
would exhibit generally weaker output growth relative to the United States than would
negative growth groups. We would also expect these relatively buoyant growth conditions to be reflected in interest rates, so that positive growth countries would generally
offer relatively lower lending rates.
We find that some of the attributes of positive and negative growth groups conform
to expectations (Table 2). The asset quality of the positive growth group of banks was
superior to that of the negative growth group. For branches and agencies and (to a lesser
extent) the subsidiaries of foreign banks, the ratio of problem loans to total corporate
loans was higher for banks from negative growth countries than for banks from positive
growth countries.
Contrary to expectations, however, the average capital ratio of the parents of branches and agencies of banks from positive growth countries was lower than that of the parents of branches and agencies from negative growth countries. A similar pattern was
observed for the subsidiaries: the average capital ratio of the subsidiaries of banks from
positive growth countries was lower than that of the subsidiaries of banks from negative
growth countries. In a later section, we note that even after controlling for other factors,
this pattern still holds for the subsidiaries. The similarity between branches and agen-
Chart 2: Commercial and Industrial Loans by Branches and Agencies of Foreign Banks
154
June 30, 1990 = 100
300
H-ll
1985
H-l H-ll
1986
H-l H-ll
1987
H-l H-ll
1988
H-l H-ll
1989
H-l H-ll
1990
H-l H-ll H-l
1991
1992
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Notes: Banks are grouped by whether commercial and industrial loans from all the banks in
each country grew after the start of the recession. Tick marks signify the end of the period.
Percents indicate share of loans by all countries as of the end of September 1992.
Chart 3: Lending to United States by Japanese Banks
Billions of dollars
80
Sources: Federal Financial Institutions Examination Council, Reports of Condition; Asset
Sales Report.
Note: Bar on the left shows C&l loans as of June 30, 1990. The bar on the right shows the
amount of loans as of September 3 0 , 1 9 9 2 , and asset-backed CP as of mid-1993.
Chart 4: Bank Lending to U.S. Corporations
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
Sources: Bank for International Settlements; Federal Financial Institutions Examination
Council, Reports of Condition; Federal Reserve Form 2502; Federal Reserve Form 2951;
Federal Reserve Bulletin, Statistical Table 4.3; FRBNY staff estimates.
Note: Tick marks refer to the end of the year.
155
Causes and Consequences
Table 2: Attributes of Positive and Negative Growth Countries
Weighted Averages
156
Bad Loan
Proportion
Average
Average
Capital Ratio
(Percent)
(Percent)
Logarithmic
Growth in GDP
Difference in
Interest Rate
(Percent)
Branches and agencies
6.34
8.53
0.011
7.17
8.53
0.003
Australia
Canada
7.46
7.71
10.40
8.81
0.006
-0.020
-8.25
France
Germany
6.89
8.03
7.46
8.65
0.015
0.056
-0.25
-1.87
Hong Kong
5.43
Positive growth
(Regression)
Israel
5.08
Korea
7.67
Netherlands
6.06
Switzerland
2.65
Negative growth
6.47
(Regression)
Italy
Japan
Spain
United Kingdom
-1.46
-1.91
-2.75
0.316
0.041
-1.75
11.40
0.018
2.08
8.71
0.042
1.26
6.48
8.71
0.043
1.29
5.23
7.06
-3.68
6.11
3.02
8.79
0.016
0.047
11.91
0.035
1.98
-5.94
22.16
9.05
-0.01
-4.00
Subsidiaries
Positive growth
4.59
9.41
-5.94
(Regression)
6.36
10.50
0.035
0.004
Spain
6.36
7.91
10.50
0.035
0.026
0.025
-0.020
0.015
0.032
-5.94
Negative growth
(Regression)
Canada
France
Hong Kong
Israel
Italy
Japan
6.83
3.07
3.51
13.1
15.08
6.16
9.89
9.76
11.57
7.94
-5.94
-0.34
-0.34
-2.75
-0.25
5.80
3.81
8.36
11.27
12.12
10.68
13.94
Netherlands
2.45
10.97
0.041
-1.75
Switzerland
3.69
10.90
10.60
0.018
2.08
7.23
-0.01
-4.00
Korea
United Kingdom
0.016
0.047
-3.68
1.98
Sources: Federal Financial Institutions Examination Council, Reports of Condition and Income; International Banking and
Credit Analysis; Bank for International Settlements; International Monetary Fund.
Notes: All averages are weighted by the amount of commercial and industrial loans outstanding for each bank as of June
30,1990. The aggregate figures for difference in interest rates and growth in GDP are weighted by the amount of C&l loans
for each country. The bad loan proportion for all banks includes loans not accruing or more than ninety days past due. The
proportion for subsidiaries also includes net charge-offs. The capital ratio for branches and agencies is that of the parent
bank, while the capital ratio for subsidiaries is that of the actual subsidiary. The bad loan proportions shown are as of the
end of 1991, while the capital ratios and difference in interest rates (U.S. capital ratios and differences in interest rates [U.S.
prime rate - foreign country lending rate]) are shown for the end of 1990. The logarithmic growth in GDP is over 1990. A
country is identified as positive growth or negative growth on the basis of its banks' growth in lending from June 30,1990, to
September 30,1992. All banks from a particular country are assigned to the same category, irrespective of the growth in
loans from a particular bank.
cies on the one hand and subsidiaries on the other is surprising since subsidiaries are
thought to resemble domestic banks and capital strength has been shown to be an important determinant of loan growth in the case of U.S.-owned banks.
Demand factors in the home country appear to be important in determining U.S.
lending by branches and agencies but not that by subsidiaries. The group of branches
and agencies exhibiting negative credit growth had a relatively higher home country
output growth than did the positive growth group. This finding suggests that higher output growth in the home country led to some reallocation of the portfolio away from the
United States and probably toward the home country. Conversely, the negative growth
group of subsidiaries had relatively lower home country growth, suggesting that home
country demand was not a strong influence on lending by the subsidiaries.
The relative differences between interest rates in negative and positive growth countries also seem puzzling. Banks from countries that had lending rates higher than those
in the United States lent more in the United States regardless of the institutional form of
the banks. This result does not hold, however, when other determinants of U.S. lending
are controlled for in a later section.
Comparing Foreign and Domestic Banks
Comparing the mean attributes of the different types of banks, we find little evidence
that supply-side factors are significant determinants of foreign bank lending in the United States. In particular, if we compare U.S. banks with the group of foreign banks from
countries that continued to extend credit, we observe that the difference in growth is not
supported by the differences in asset quality (Table 3). In most cases even the signs are
wrong. That is, the group of banks that grew rapidly had the poorer portfolio. Indeed,
the most striking negative result is that branches and agencies of foreign banks increased their lending significantly more than the smaller domestic banks although the
asset quality of their parents was significantly worse than that of these domestic banks.
Moreover, while the asset quality of parent banks did not differ significantly from that
of the larger domestic banks, the growth in their branches and agencies was significantly
higher. Only in one comparison—large U.S. banks and subsidiaries of foreign banks—
was the sign in the expected direction: that is, subsidiaries that grew faster than the large
domestic banks also had the better asset quality. Nevertheless, neither the loan growth
difference nor the asset quality difference was significant.
Slicing the same groups of banks according to their capital strength leads to similar
conclusions. Where the differences in loan growth were significant, differences in asset
quality were not and vice versa. Indeed, very often the signs were contrary to expectations. That is, even after capital strength was held constant, the more rapidly growing
foreign banks often had the worst asset quality. In particular, branches and agencies of
those positive growth banks that were more weakly capitalized grew significantly faster
than the more weakly capitalized U.S. banks—both small and large (Table 4). These
parent banks, however, had a weaker balance sheet than either group of domestic banks.
Similarly, the branches and agencies of the better capitalized parents grew more than either the small or large domestic banks although their asset quality was worse. Note,
however, that in three of the four comparisons just made, the loan growth differed significantly for the two groups but the asset quality did not, and in the remaining comparison neither difference was statistically significant. These comparisons suggest that the
perverse result that we observed earlier is less strong once we control for capital
strength.
157
Causes and Consequences
Comparing Foreign Banks
Conclusions similar to those above emerge when we compare foreign banks grouped according to both institutional type and capital strength. In general a close relationship
between loan growth, asset quality, and capital strength is not borne out (Table 5). In
particular, we find a significant difference in growth between the more weakly capitalized branches and agencies that showed positive growth and those that showed negative
growth, but no significant difference in asset quality between these two groups. A perverse result that emerges is that the negative growth branches and agencies with a weaker capital base contracted less than those with the stronger capital base. By contrast, the
behavior of well-capitalized branches and agencies is consistent with expectations—
banks with the better asset quality also had significantly stronger loan growth.
The descriptive analysis of the influence of capital strength and of asset quality can
control for other factors only in a very limited way. To overcome this limitation, we will
Table 3: Comparison of Mean Attributes of Banks
Difference in Weighted Means of the Two Groups: Domestic Banks minus Foreign Banks
(Foreign Banks from Countries with an Expanding C&l Portfolio)
Logarithmic Growth of Commercial and Industrial Loans from 1990 to 1992
Domestic Banks
Foreign Banks from Countries with
an Expanding C&l Portfolio
Subsidiaries
Branches and agencies
Small Banks
Large Banks
-28.50
(-1.16)
-23.06
(-0.93)
-63.05
(-4.54) ***
-57.51
(-4.03) ***
Bad Loan Proportion as of the End of 1991 (Percent of Loans Past Due 90+ Days or Not Accruing)
158
Foreign Banks from Countries with
an Expanding C&l Portfolio
Subsidiaries
Branches and agencies
Domestic Banks
Small Banks
Large Banks
-0.88
(-0.47)
1.12
(0.56)
-3.87
(-2.57) ***
-1.77
(-1.14)
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Notes: Comparison is based on bank-specific (as opposed to country-specific) data. A bank is
categorized according to aggregate commercial and industrial loan growth by all banks from its
home countries. Numbers in parentheses are t-statistics. The cutoff for big banks is $1 billion
in assets as of the end of 1991. Variances are assumed to be not equal. The means are
weighted by the amount of commercial and industrial loans outstanding at the start of the
recession. A constant sample of banks is used in computing the data for this table.
* Means are significantly different at the 10 percent level.
** Means are significantly different at the 5 percent level.
*** Means are significantly different at the 1 percent level.
Table 4: Comparison of Mean Attributes of Banks
Difference in Weighted Means of the Two Groups: Domestic Banks minus Foreign Banks
(Foreign Banks from Countries with an Expanding C&l Portfolio)
Logarithmic Growth of Commercial and Industrial Loans from 1990 to 1992
Foreign Banks
from Countries
with an
Expanding
C&l Portfolio
Domestic Banks
Small Banks
Low Capital
Ratio
High Capital
Ratio
Large Banks
Low Capital
Ratio
High Capital
Ratio
Subsidiaries
Low capita! ratio
-53.53
(-1.10)
-73.68
(-1.51)
13.41
(1.19)
4.67
(0.43)
High capital ratio
Branches and agencies
Low capital ratio
-90.40
(3.29) ***
-70.16
(-2.57) **
-65.37
(-1.90)*
High capital ratio
-56.62
(-1.64)
Bad Loan Proportion as of the End of 1991 (Percent of Loans Past Due 90+ Days or Not Accruing)
Foreign Banks
from Countries
with an
Expanding
C&l Portfolio
Domestic Banks
Small Banks
Low Capital
Ratio
High Capital
Ratio
Large Banks
Low Capital
Ratio
High Capital
Ratio
Subsidiaries
Low capital ratio
2.79
(4.55) ***
2.77
(3.10)***
-2.44
(-0.63)
High capital ratio
-1.52
(-0.39)
Branches and agencies
Low capital ratio
High capital ratio
-3.47
(-1.05)
-3.48
(-1.04)
-2.35
(-0.98)
-1.43
(-0.59)
Sources: Federal Financial Institutions Examination Council, Reports of Condition; International Banking and Credit Analysis.
Notes: Comparison is based on bank-specific (as opposed to country-specific) data. A bank is
categorized according to aggregate commercial and industrial loan growth by all banks from its
home countries. Numbers in parentheses are t-statistics. The cutoff for big banks is $1 billion
in assets, and the cutoff for the high capital ratio is 9 percent. The cutoffs are for the end of
1990 for capital ratios, and for the end of 1991 for assets. Variances are assumed to be not
equal. All subsidiaries and domestic banks are included in the table, but only branches and
agencies from countries whose parent banks disclose capital ratios are included. The means
are weighted by the amount of commercial and industrial loans outstanding at the start of the
recession. A constant sample of banks is used in computing the data for this table.
* Means are significantly different at the 10 percent level.
** Means are significantly different at the 5 percent level.
*** Means are significantly different at the 1 percent level.
159
Causes and Consequences
Table 5: Comparison of Mean Attributes of Country Groupings of Banks
Difference in Weighted Means of the Two Groups (Column minus Row)
Logarithmic Growth of Commercial and Industrial Loans from 1990 to 1992
Branches and Agencies
Positive Growth
Negative Growth
Low Capital Ratio High Capital Ratio Low Capital Ratio High Capital Ratio
Branches and agencies
Low capital ratio
Negative growth
-54.38
(-2.24) *
80.31
(2.05) **
41.59
(0.61)
Positive growth
-80.31
(-2.05) **
High capital ratio
Positive growth
-176.26
(-2.92) **
-41.59
(-0.61)
Negative growth
176.26
(2.92) **
54.38
(2.24) *
Subsidiaries, negative growth
Low capital ratio
High capital ratio
11.04
(0.86)
-37.06
(-1.47)
Bad Loan Proportion as of the End of 1991 (Percent of Loans Past Due 90+ Days or Not Accruing)
160
Branches and Agencies
Positive Growth
Negative Growth
Low Capital Ratio High Capital Ratio Low Capital Ratio High Capital Ratio
Branches and agencies
Low capital ratio
-0.63
3.49
Negative growth
(0.10)
(0.36)
-2.91
0.63
Positive growth
(-0.94)
(-0.10)
High capital ratio
2.91
7.03
Positive growth
(0.94)
(0.91)*
-7.03
-3.49
Negative growth
(-0.91)*
(-0.36)
Subsidiaries, negative growth
-1.32
Low capital ratio
(-0.14)
4.45
High capital ratio
(0.55)
Sources: Federal Financial Institutions Examination Council, Reports of Condition; International Banking and Credit Analysis.
Notes: Comparison is based on bank-specific (as opposed to country-specific) data. A bank is categorized according to
aggregate commercial and industrial loan growth by all banks from its home country. The means are weighted by the
amount of commercial and industrial loans outstanding at the start of the recession. Numbers in parentheses are t-statistics. The cutoff for the high capital ratio category is 9 percent as of the end of 1990. Variances are assumed to be not
equal. All subsidiaries are included in this table, but only branches and agencies from countries that disclose parent bank
capital ratios are included.
* Means are significantly different at the 10 percent level.
" Means are significantly different at the 5 percent level.
" * Means are significantly different at the 1 percent level.
introduce an equation that controls for many factors simultaneously, but first we offer a
simple capital adequacy model.
IV. Capital Adequacy
This section offers a direct test of the hypothesis that capital strength determined the rate
of expansion of foreign banks. Recent literature on the credit crunch has argued that
capital adequacy contributed importantly to the downsizing of balance sheets by domestic banks in recent years.12 The onset of the BIS era has drawn attention to the risk-adjusted capital ratios at banks. In an effort to establish uniform capital standards across
banks from different countries, the Basle capital regulations have stipulated that banks
attain a risk-adjusted capital ratio of at least 8 percent. To achieve these standards,
banks with low ratios can either raise capital or, in the face of difficult stock market conditions, reduce growth.
The simple reduced-form model that we test takes the following form:
where
C = log change in C&I loans,
k = risk - adjusted capital ratio,
and the subscripts
/ = bank /
t = time t.
A few points should be noted. First, both the capital ratio and loan growth are bankspecific variables. Second, we run two sets of regressions, the first using the actual capital ratio, labeled CAPRATIO, and the second using a dummy variable, CAPDUM, that
is equal to one if the risk-adjusted capital-asset ratio is greater than or equal to nine.13
Third, the growth in commercial and industrial loans is computed over the recession—
1990-11 through 1991-1—and the end-1990 capital ratio is used. The regressions thus
represent a pure cross-section analysis. Finally, we run a set of regressions for both
branches and agencies and subsidiaries using country dummies to control for countryspecific factors.
Results of these regressions, reported in Table R-l, confirm the view that capital adequacy was not a significant determinant of foreign bank loan growth. The capital ratio
is significant and of the right sign in only one regression for the branches and agencies
that uses the actual capital ratio and does not control for country-specific factors. This
regression, however, has very low explanatory power. Once country dummies are in12
See, for example, Richard F. Syron, "Arc We Experiencing a Credit Crunch?" Federal Reserve Bank of
Boston, New England Economic Review, July/August, pp. 3-10; Joe Peek and Eric Rosengrcn, "The Capital Crunch in New England," Federal Reserve Bank of Boston, New England Economic Review, May/June
1992, pp. 21-31; Ronald Johnson, 'The Bank Credit 'Crumble'"; and Ben S. Bcrnankc and Cara S. Lown,
"The Credit Crunch," Brookings Papers on Economic Activity, 2:1991, pp. 205-47.
13
We chose nine as the cut-off rather than the BIS-stipulated ratio of eight since almost all banks would otherwise have been in the "high capital ratio" group. We felt that this categorization would not have represented the vulnerability of the capital ratio of many banks to market conditions.
161
Causes and Consequences
troduced in the regression, the explanatory power of the regression improves considerably, but the importance of capital strength disappears. These findings suggest that
capital strength served as a relatively weak proxy for country-specific factors. In the
next section, we explore which country-specific factors mattered most.
V. Country-Specific Factors
This section examines a reduced-form model of foreign bank lending. In addition to the
effect of capital strength on credit extended by foreign banks, we examine the effect of
asset quality by controlling for several factors. These factors include foreign direct nonbank investment in the United States, the price-earnings ratio, market capitalization of
the home-country14 stock market, growth in real GDP in the United States as well as in
the home country, the difference in lending rates between the United States and the
home country, bank loans to all U.S. nonbanks booked in the home country, and last period's growth in commercial and industrial loans of the bank in question. As in the previous section, we run the regressions separately for the two institutional types—
branches and agencies, and subsidiaries. In addition, we extract the positive growth
group to identify the determinants of lending of the group of foreign banks that differed
in their credit behavior from U.S.-owned banks.
In addition to capital strength as defined in the last section, we examine the role
played by credit quality. Poor asset quality has largely resulted from a combination of
bank exposure to real estate loans and rapidly declining real estate values. In our regres14
Home country refers to the country of domicile of the parent of the foreign bank.
Table R-1: Regression over Commercial and Industrial Loan Growth
162
Branches and Agencies
Without Country
Dummies
INTERCEPT
-0.03
(-1.0)
(3.3)
0.03
(1-2)
0.02
(0.3)
0.04
(0.6)
Without Country
Dummies
-0.04
(-1.6)
•0.75 ***
0.08 ***
(3.2)
CAPRATIO
CAPDUM
With Country
Dummies
Subsidiaries
With Country
Dummies
C.01
(5)
-0.00 *
(-1.8)
-0.01 **
(-2.5)
-0.04
(-1.0)
•0.00
(0.1)
F
0.3
10.4***
6.0 ***
6.1 ***
0.0
3.4*
3.4 ***
3.7 ***
R-Square
0.00
0.04
0.31
0.32
0.0
0.02
0.50
0.53
Adj. R-Square
-0.00
0.04
0.26
0.26
-0.0
0.02
0.36
0.38
INTERCEPT
CAPRATIO
CAPDUM
Intercept
Actual Capital Ratio
Dummy variable=1 if capital ratio is greater than or equal to 9, else variable=0
Notes: Capital ratio is for the end of 1990. Commercial and industrial loan growth is from
June 30, 1990, to March 31, 1991.
sions, we use the percentage of loans not accruing or past due ninety days or more,
BADPROP, to proxy for the extent of bad loans in each bank's portfolio. We expect a
negative sign for the coefficient of BADPROP because the more numerous the bad
loans, the poorer the asset quality and the slower we expect loans to grow.
The literature on foreign bank activity in the United States has frequently tied the
growth of foreign banks to the servicing of home country clients.15 This servicing of
home clientele would suggest a positive sign on the coefficient of foreign direct investment capital flows, FDIFLOW, to the United States from the home country of the foreign bank.
It has been argued elsewhere that the willingness of foreign investors to accept lower
returns than do U.S. investors has probably been the most important determinant of foreign bank growth in the United States.16 To examine the effect of differences in cost of
equity on the growth of foreign banks, we include the price-earnings ratio of the stock
market, PERATIO, in the home country of the foreign bank. The lower the price-earnings ratio, the lower the cost of equity, and the greater we can expect the foreign bank
expansion to be. Thus we expect the sign on this coefficient to be negative.
Wealth effects may also affect foreign bank activity. For home country investors,
more wealth should create greater demand for all instruments, assuming the investor's
portfolio is appropriately diversified, and should lead to greater direct foreign investment, including investment in foreign banking. It has been pointed out that "greater
fund availability" by foreign banks could account for their growing presence. l7 It is possible that wealth effects underlie this argument. These wealth effects would suggest a
positive coefficient on market capitalization or MARKCAP.
A demand factor that has been included in studies on domestic banks is the growth
in real gross domestic product (GDP). Stronger domestic growth, termed USGDP in our
regressions, has been found to increase the demand for credit from domestic banks. By
the same token, the growth in real GDP of the home country of the foreign bank could
be an important demand factor in determining credit in that country. It follows that all
things being equal, the slower the growth in real GDP in the foreign bank's home country, FORGDP in our regressions, the more rapid the bank's expansion of credit abroad,
in particular credit to the United States.18
Differences in lending rates between the home country of the foreign bank and the
United States, adjusted for expected movements in the exchange rate, are also likely to
affect foreign bank growth in the United States. Differing GDP growth rates for the two
15
At least as early as 1979, a study by the General Accounting Office suggested that among the reasons for
the influx of foreign banks into the United States was the "following by foreign banks of foreign business
to the United States." See Comptroller General of the United States, General Accounting Office (1979, p.
3). Also see Clifford A. Ball and Adrian E. Tschoegl, "The Decision to Establish a Foreign Bank Branch or
Subsidiary: An Application of Binary Classification Procedures," Journal of Financial and Quantitative
Analysis, vol. 17, no. 3 (September 1982), pp. 411-24; Lawrence G. Goldberg and Anthony Saunders, "The
Determinants of Foreign Banking Activity in the United States," Journal of Banking and Finance, vol. 5
(1981), pp. 17-32; Charles W. Hultman and Randolph McGcc, "Factors Affecting the Foreign Banking
Presence in the United States." Journal of Banking and Finance, vol. 13, no. 3 (July 1989), pp. 383-96.
16
McCaulcy and Scth, "Foreign Bank Credit to U.S. Corporations." See also Steven A. Zimmcr and Robert
N. McCaulcy, "Bank Cost of Capital and International Competition," Federal Reserve Bank of New York
Quarterly Review, vol. 15 (Winter 1991), pp. 35-59.
17
Gary C. Zimmerman, "The Growing Presence of Japanese Banks in California," Federal Reserve Bank of
San Francisco Economic Review, Summer 1989, pp. 3-17.
18
We are assuming that, at least over the period of analysis, foreign banks face an inelastic supply of capital.
163
Causes and Consequences
164
countries constitute one factor explaining the interest rate differentials but not the only
factor. We therefore include in the regressions differences in lending rates, DIFFINT,
adjusted for exchange rate changes, EXCHANGE. It should be noted that we do not
restrict the coefficient to be identical for the rate differential and exchange rate changes.
Foreign banks lend to U.S. corporations both by booking loans onshore and by booking them offshore.19 In examining the factors determining growth in loans booked by
foreign banks onshore, we would have to control for the loans they booked offshore.
The two types of loans could conceivably be substitutes or complements. If the loans
are substitutes, foreign banks would book fewer onshore loans if their offshore book was
increasing rapidly. This situation would arise if, say, it was more cost effective to borrow abroad and lend offshore. If the loans are complements, these banks would be increasing their onshore book at the same time that their offshore book was increasing.
Since bank-specific data are not available for offshore loans, we include in our regressions all bank flows to nonbanks in the United States from the home country of the bank
in question—BISFLOW. Finally, to capture a possible adjustment factor, we include the
past period's growth in commercial and industrial loans by the foreign bank—OLDCIGRO. This variable incorporates the idea that foreign banks are seeking a "target"
market share. The faster the growth in previous periods, the slower the current growth
has to be to adjust to this desired share level.
Incorporating the above variables in Equation 1, we test the following reduced-form
equation:
Q, =/(*,„ qiP Wj, COEj, FDljp GDPtUS, GDP{ (itUS -ij), el OUP Cit_,)t
+ - +
+
+
+
- ?
where
C = log change in C&I loans
k = risk - adjusted capital ratio (or a dummy)
q = percent of total loans 90 days past due and not accruing (and also
charged off, in the case of subsidiaries)
VV= total stock market capitalization
p = price - earnings ratio of the stock market
FDI = foreign direct investment capital flows into the United States
GDP = log change in real GDP
i = bank lending rate
e = exchange rate (foreign currency per U.S. dollar)
OL = cross-border bankflows to U.S. banks
and the subscripts
! =: bank i
j =
home country j of bank i
t=
time t
and the superscript
j
19
=
home country of bank i.
McCauley and Seth, "Foreign Bank Credit and U.S. Corporations."
VI. Results
The results of the regressions using this model are reported in Table R-2. Although only
about 10 percent of the variance is explained, some interesting results emerge. Factors
determining foreign bank lending differ across institutional types. Although supplyside factors are generally not very important, capital strength does appear to have been
important in determining lending by the positive growth branches and agencies. By
contrast, we get a negative but insignificant sign on the capital coefficient for subsidiaries—a surprising finding given the similarity of subsidiaries to domestic banks. Reported asset quality does not emerge as significant in either case.
U.S. demand was important in determining lending by foreign banks but in an unexpected way. The coefficient on U.S. GDP was negative though significant only in the
case of branches and agencies. The negative sign suggests that in recessionary circumstances, foreign banks actually increased their lending to U.S. corporations. Earlier
studies have shown that domestic banks tightened lending in response to the weak state
of the U.S. economy. The market niche exposed by this credit slowdown appears to
have been filled by foreign bank credit, probably in a bid by foreign banks to increase
their market share in the United States. This inference is supported by surveys that have
polled treasurers of large U.S. corporations.20 The responses indicate that U.S. corporations did indeed shift some of their credit demand to foreign banks as these banks aggressively priced their loans, probably in order to increase their market share.
The positive growth branches and agencies seem to have diverted some of their loans1
booking to the United States from overseas branches, including those in offshore centers.
This diversion of loans to the United States was to be expected once reserve requirements,
considered a primary determinant of the rapid growth in offshore loans, were reduced to
zero. In addition, higher U.S. lending rates relative to home country rates adjusted for
exchange rate expectations also provided an impetus to lending by this group.
Market share considerations may also have prompted lending by the small group of
subsidiaries that increased credit during the recent recession. The only variable that has
a significant coefficient is the "adjustment factor variable." The lower the previous period's growth, the faster this group of subsidiaries grew. One interpretation is that if the
subsidiaries had been expanding slowly in one period, their market share could only be
made up in the next period by expanding more rapidly.
We offer some cautions about our findings in this paper. The first is obviously the
modest nature of these results, and the second is that the group of subsidiaries that increased credit and for which the results are reported here is extremely small. Two additional cautions apply particularly to our results for the branches and agencies of foreign
banks, which form roughly two-thirds of all foreign banks in the United States. First,
the assets that we are looking at are only part of the global assets of the larger entity to
which the branch or agency belongs. Second, the asset quality may be reported according to accounting standards that are not uniform across countries.
VII. Conclusions
Foreign banks as a group expanded lending to U.S. corporations even as domestic banks
were cutting back. Across foreign banks, however, differences in lending behavior are
evident. Most subsidiaries and two-thirds of the branches and agencies of foreign
banks, by value of commercial and industrial loans (just under one-half by number of
20
These surveys are conducted annually by Greenwich Associates in Connecticut.
165
Causes and Consequences
Table R-2: Determinants of Foreign Banks' Commercial and Industrial Loan Growth
Banks from Positive Growth Countries
166
Branches and Agencies
Subsidiaries
INTERCEP
0.21
(2.5) ***
-0.27
(-0.2)
USGDP
-1.74
(-1.7)*
-0.13
(-0.1)
FORGDP
-0.39
(-0.4)
-0.53
(-0.1)
CAPDUM
0.08
(2.4)"
-0.01
(-0.2)
DIFFINT
0.02
(2.3) **
0.01
(0.3)
FDIFLOW
-0.00
(-1.5)
0.00
(0.3)
MARKCAP
-0.00
(-1.5)
-0.00
(-0.2)
PERATIO
0.00
(0.0)
0.02
(0.2)
-0.52
(-1.8)*
-0.06
(-0.2)
BISFLOW
-0.00
(-2.5) ***
0.00
(0.3)
BADPROP
-0.00
(-1.3)
-0.01
(-1.1)
OLDCIGRO
-0.05
(-1.4)
-0.28
(-3.4) ***
(6.8) ***
(1.7)*
R-Square
0.083
0.119
Adj. R-Square
0.070
0.049
Variables
EXCHANGE
F
Variable Label
INTERCEP
USGDP
FORGDP
CAPDUM
DIFFINT
FDIFLOW
MARKCAP
PERATIO
EXCHANGE
BISFLOW
BADPROP
OLDCIGRO
Intercept
Log change in U.S. real GDP
Log change in foreign country real GDP
Variable is 1 if capital ratio is greater than or equal to 9, else variable is 0. For the
branches and agencies regression, the capital ratio of the parent bank is used in all
cases. For the subsidiaries regression, the capital ratio of the subsidiary as of the end
of 1990 is used for all time periods.
U.S. prime rate - foreign lending rate
FDI capital inflows to U.S.
Market capitalization for stock market of country
P/E ratio for stock market of country
Log change in exchange rate
Capital flows to U.S. nonfinancial firms from the country
Percentage of loans not accruing or past due 90 days or more
Lagged 1-quarter value of C&l loan growth
Notes: The branches and agencies regression begins in 1989 or 1990 (depending on the bank), while the subsidiaries regression begins at the end of 1987. In all cases, the regressions end in the fourth quarter of 1991.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
banks), actually decreased lending in tandem with U.S.-owned banks. Nevertheless, the
growth in loans from the group of foreign banks that increased credit outweighed the
contraction in loans from the group of banks that reduced credit. Moreover, if we include in our consideration the banks' sponsorship of asset-backed commercial paper
programs, even those foreign banks that cut back credit may have increased their total
commitments to U.S. corporations.
The evidence indicates that the supply-side factors found to be important in determining domestic bank corporate loan contraction have not been equally important determinants of foreign bank lending. In particular, capital strength has affected the group
of branches and agencies that increased credit but not subsidiaries that did the same. Reported asset quality did not appear significant in explaining lending by either group.
In the case of branches and agencies, banks that increased credit were motivated by
the desire to increase their U.S. market share and by the existence of better lending rates
in the United States than in their home country (adjusted for exchange rate expectations). Finally, some of the increased credit probably reflected foreign banks' moves to
book loans onshore that they had previously booked offshore as reserve requirements in
the United States were reduced to zero.
Market share considerations probably also explain the lending behavior of the small
group of subsidiaries that increased credit. The slower the credit growth in one period,
the faster the growth in the next. One might infer from this pattern that these foreign
banks desired to increase market share. A credit slowdown in one period may thus have
been followed in the next by aggressive credit increases designed to give subsidiaries a
higher share.
Because we consider only a partial set of assets for branches and agencies of foreign
banks and because accounting standards vary across countries, we cannot conclude that
supply-side factors were unimportant to foreign banks. Nevertheless, the differing response of the foreign banks was undoubtedly useful: U.S. borrowers may have been able
to obtain credit that would not have been available had the banking system been more
homogeneous.
167
Causes and Consequences
Appendix
Appendix Table: Growth of Commercial and Industrial Loans by Foreign Banks
Categorized by Country (Percent)
168
Branches and Agencies
Subsidiaries
Growth from
1990-11
to
1992-111
Share of
Subsidiary
Loans
1992-111
0.00
-27.69
0.0
-79.85
0.04
-30.95
0.1
Bahrain
-73.34
0.02
Chile
-66.17
0.00
Argentina
-60.77
0.0
Cayman Islands
-51.95
0.01
New Zealand
-49.12
0.02
Yugoslavia
-45.93
0.10
-24.62
0.1
India
-44.38
0.12
35.71
0.1
Spain
-44.17
0.38
194.37
10.1
Panama
-42.14
0.00
U.K.
-37.61
1.03
-13.93
17.0
Growth from
1990-11
to
1992-111
Share
of B&A
Loans
1992-111
Peru
-93.24
0.00
U.A.E.
-82.12
Denmark
Brazil
-37.22
0.11
-22.60
1.9
Ireland
-32.39
0.03
-28.65
2.4
Puerto Rico
-27.31
0.12
Kuwait
-25.94
0.02
Greece
-25.31
0.01
-38.63
0.3
-17.11
0.7
-37.51
0.5
-0.52
39.3
Italy
-24.43
3.95
Belgium
-15.61
0.05
Venezuela
-15.23
0.01
Malaysia
-13.95
0.17
Jordan
-9.53
0.00
Japan
-2.31
55.24
Guam
0.67
0.00
Singapore
2.17
0.04
Philippines
7.03
0.01
-26.64
0.2
Indonesia
8.23
0.15
-20.31
0.7
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Notes: Only countries that were included in the analyses are shown. Excluded countries
did not have a branch, agency, or subsidiary with a positive amount of loans on June 30,
1990, or September 30, 1992. Iran and Saudi Arabia were also excluded since they had
very small amounts of loans in 1990. An empty cell means that the foreign country did not
maintain that kind of foreign bank during the 1990-92 period. Boldface type denotes positive growth.
Appendix Table: Growth of Commercial and Industrial Loans by Foreign Banks (Continued)
Categorized by Country (Percent)
Branches and Agencies
Subsidiaries
Growth from
1990-11
to
1992-111
Share of
Subsidiary
Loans
1992-111
1.11
-30.62
2.6
54.59
1.41
-17.00
0.3
Pakistan
76.35
0.01
China
80.61
0.16
Australia
90.94
0.60
Portugal
94.08
0.25
-51.07
0.0
Taiwan
99.16
0.62
-40.43
0.1
Egypt
104.55
0.00
Mexico
110.14
0.09
-43.43
0.2
Finland
115.75
0.69
Norway
127.31
0.44
Canada
128.91
9.46
-16.08
5.7
Hong Kong
156.61
1.01
-37.65
6.4
Growth from
1990-11
to
1992-111
Share
ofB&A
Loans
1992-111
Austria
15.45
0.06
Germany
16.13
1.67
Thailand
25.96
0.07
Israel
29.58
Korea
France
206.22
7.94
-31.52
0.7
Netherlands
220.49
5.47
-57.47
0.6
Switzerland
279.91
6.80
-12.32
0.0
Sweden
435.19
0.35
-62.32
0.0
El Salvador
Other Western Europe
-53.12
1.1
Multiple countries
-45.56
3.7
Bermuda
-39.08
0.0
Columbia
-34.91
0.1
Ecuador
-29.30
0.2
Other Caribbean
57.73
0.1
Netherlands Antilles
68.61
4.6
Dominican Republic
156.73
0.0
Source: Federal Financial Institutions Examination Council, Reports of Condition.
Notes: Only countries that were included in the analyses are shown. Excluded countries
did not have a branch, agency, or subsidiary with a positive amount of loans on June 30,
1990, or September 30, 1992. Iran and Saudi Arabia were also excluded since they had
very small amounts of loans in 1990. An empty cell means that the foreign country did not
maintain that kind of foreign bank during the 1990-92 period. Boldface type denotes positive growth.
169
Causes and Consequences
170
Nonbank Lenders and the Credit Slowdown
by Richard Cantor and Anthony P. Rodrigues1
Many analyses of the recent slowdown in business credit growth have focused on the
decline in commercial and industrial bank loans. In contrast, we examine in this paper
the behavior of nonbank business credit, which consists primarily of corporate bonds,
commercial paper, and loans from finance companies. Credit availability from nonbank
sources began to decline at year-end 1989, particularly for small firms and other borrowers without strong credit ratings. Around that time, the junk bond market collapsed, real
estate problems and a recession emerged in New England, and defaults on bonds and
commercial paper soared. Over the next three years, the actions of rating agencies, private investors, and regulators encouraged all financial intermediaries—banks, finance
companies, and insurance companies—to improve their asset quality, strengthen their
capital bases, and limit their balance sheet growth. The relatively smooth functioning
of the capital markets during this otherwise turbulent period, however, did reduce the
impact of the credit crunch by allowing borrowers with strong credit ratings to obtain
funds at reasonable rates. Moreover, the market's ready acceptance of asset-backed securities enabled many hobbled financial intermediaries to reduce their funding and capital needs without dramatically cutting back on loan origination.
During the 1990-91 credit crunch period, nonbank business lending did grow faster
than bank loans, but the main causes of the aggregate slowdown do not appear to be limited to problems specific to the banking system. Nonbank credit had already been growing faster than bank credit since the mid-1980s. During the recent crunch period, bank
and nonbank business credit growth rates in fact declined by roughly the same amounts
and the timings of their decelerations were also very similar. The causes of the credit
slowdown must, therefore, include factors that affected nonbank as well as bank sources
of credit.
Our paper emphasizes factors that led to a decline in nonbank credit availability, although cutbacks in demand and supply both likely contributed to the credit slowdown
1
We are particularly grateful to Gary Haberman for his comments and also owe thanks to Akbar Akhtar,
Dick Davis, Benjamin Friedman, Cara Lown, Trish Mosser, Mitch Post, Charles Steindel, and John Wenninger. Excellent research assistance was provided by Kin Cheng, Jeanctte Donato, Bill LeCates, and Ted
Manton.
171
Causes and Consequences
and its similarity in appearance across different sources of credit. Credit demand undoubtedly fell due to the weakness in economic activity and to borrower efforts to reduce indebtedness. Contracting credit supply, however, may also explain a substantial
portion of the parallel decline in bank and nonbank credit. Our study suggests that capital markets and nonbank financial institutions experienced shocks and stresses that induced a decline in credit supply similar to that observed in the banking industry.2
In Section I, we present a summary of our view of the events leading up to the current
period of credit stringency, emphasizing developments affecting nonbank sources of
business credit and making some comparisons to previous crunches. Section II presents
time series evidence on the slowdown in business credit and compares the recent behavior of various nonbank sources of credit to their behavior in previous recessions and
crunches. In contrast to past episodes, we find little evidence of a shift from bank to
nonbank sources of funds during the current slowdown. We then employ a simple
econometric model to test whether the slow growth rates of bank and nonbank credit
were quantitatively unusual given their historical relationships with other macroeconomic variables. Here we confirm the similarities, both in terms of magnitude and timing, between the slowdown in bank and nonbank business credit growth. Section III
reviews the recent stresses experienced by nonbank financial intermediaries and the fragility visible in parts of the organized credit markets. This evidence suggests that credit
availability from nonbank funding sources has constrained some borrowers, particularly
firms that are small or rated below-investment grade. We conclude with a review of the
role played by the asset-backed securities market in relieving some of the pressures on
intermediaries associated with the credit crunch. In the Appendix, we provide a brief
review of the major sources of business credit.
Overview of Recent Factors Affecting Nonbank Business Credit
172
The slowing of credit growth that occurred in 1966, 1969, 1974 and 1982 was prompted
by a tightening of monetary policy in response to rising inflation. On those occasions,
banks and thrifts lost deposits as market rates rose above Regulation Q interest rate ceilings.3 This disintermediation process led to a contraction in bank loans and thrift mortgages. Business borrowing from nonbank sources generally rose in the beginning of
these credit crunches, as firms substituted away from bank loans to commercial paper
issuance and other funding sources. Nonbank business lending later fell, though, as
credit demand weakened during the ensuing decline in economic activity.
The current recession was also preceded by a sharp contraction in the rate of business
borrowing which was greater in real terms than observed in previous credit crunches.
This credit slowdown appeared particularly sharp in comparison to the extremely rapid
real rates of credit growth experienced during the mid-1980s. Compared to past credit
crunches, rising interest rates played a much less significant role in the current episode,
2
3
For review of other papers on the bank credit crunch and some independent analysis as well, see see the
paper by Cara Lown and John Wenninger in this volume.
The characterization presented above of post-War credit crunches emanates from the writings of Wojnilower and Kaufman in the mid-1960s. A historical review of credit crunches and the related literature is
contained in Richard Cantor and John Wenninger, "Perspective on the Credit Slowdown," Quarterly
Review, Federal Reserve Bank of New York, Spring, 1992-93. Government officials may have also played
a direct role in past credit crunches when they pressured banks to maintain a constant prime lending rate
while restraining credit growth as part of campaign against inflation. This aspect of credit crunches has
been emphasized by Raymond Owens and Stacey Schreft in "Identifying Credit Crunches," Federal
Reserve Bank of Richmond, Review, 1992.
as short-term rates fell continuously after reaching a modest peak in early 1989. Moreover, in contrast to previous crunches, most depository institutions did not report unusual liquidity problems. Furthermore, the decline in credit growth was much more evenly
distributed across a wide variety of funding sources, including bank loans, below-investment-grade bonds, commercial paper, and finance company loans.
Beginning in 1989, a variety of factors converged to reduce the willingness and ability of both banks and nonbanks to lend to businesses, particularly lower quality credits.
The economy's long lasting expansion began to slow and the commercial real estate
market soured, weakening the balance sheets of both financial intermediaries and prospective borrowers. In October 1989, the stock experienced a mini-crash and the junk
bond market collapsed, putting an end to the M&A/LBO-related activities that had propelled bank and nonbank business credit for much of the decade.
Defaults and bankruptcies began to rise, not only among junk bond issuers, but
among the thrifts, life insurers and finance companies that had invested in junk bonds
and commercial real estate. The credit rating agencies, reflecting the concerns of investors more generally, increased their scrutiny of financial institutions and issued a rash of
rating downgrades in late 1989 and early 1990. At the same time that financial institutions were trying to improve their capital ratios and asset quality, many borrowers themselves decided to strengthen their balance sheets by paying off debt or foregoing
planned borrowing.
The financial regulators took aggressive actions to deal with the emerging credit
problems and reassure creditors of financial intermediaries. Following the shocking
commercial real estate losses announced by New England banks beginning in the Fall
of 1989, bank examiners began paying closer attention to the quality of real estate credits across the nation. In mid-1990, following some defaults in the commercial paper
market, the Securities and Exchange Commission adopted a rule that limited the amount
that money market mutual funds could invest in securities that did not carry the highest
possible credit ratings. Around the same time, following several prominent life insurer
failures and near-failures, the state insurance regulators adopted rules that discouraged
investments in below-investment grade bonds and required greater disclosure and reserves for such bonds.
By early 1990, both banks and finance companies had raised their credit standards,
widened their lending spreads, and slowed the rate of growth in business loans.4 Life
insurance companies shifted their investments away from commercial real estate mortgages, junk bonds, and private placement issues of middle market firms to higher quality
credits such as investment grade corporate bonds, treasury bonds and mortgage-backed
securities. Firms of lower credit quality found borrowing difficult in the long-term credit markets and in the commercial paper market.
Nonfinancial firms with strong credit ratings, however, had uninterrupted access to
credit in the money and capital markets—as investment-grade quality yield spreads,
both at short and long maturities, remained relatively flat throughout the recession. Junk
bond issuance, however, collapsed in late 1989 and the yield spread between above- and
below-investment-grade bonds widened dramatically shortly thereafter. Commercial
paper issuance, by companies other than those with the highest credit ratings, also
shrank in 1990. Yet the market for high quality fixed-income securities remained strong,
4
Survey evidence lhat banks lightened credit availability significantly, particularly for small firms, is presented by Kausar Hamdani, Anthony Rodrigucs, and Maria Varvatsoulis in this volume. In their annual
reports and informal discussions, finance companies executives have staled directly that they raised their
credit standards in 1990.
173
Causes and Consequences
and many financial intermediaries were able to issue asset-backed securities that relieved, in part, their balance sheet constraints and allowed them to continue lending.
II. Business Credit Growth and Historical Comparisons
In order to understand the effects of the events described in the historical narrative
above, we review in this section the quantitative behavior of various sources of business
credit before and during the current period of credit stringency. We first document the
unusually rapid real growth in business credit during the mid-1980s and highlight the
abruptness of the slowdown in credit growth that began in 1989. We then provide a
graphical comparison of the recent behavior of various components of total business
credit with their behavior in previous credit crunches. When comparing the relative
movements of bank and nonbank credit in the early stages of the credit crunches, we find
little evidence of a shift from bank to nonbank sources of funds during the current slowdown in contrast to past episodes. In a regression framework, we then confirm the similarities, both in terms of magnitude and timing, between the slowdown in bank and
nonbank credit growth.
Overview of Trends in Business Credit Growth
174
The top panel of Chart 1 displays the annual growth rate of nonfinancial business sector
liabilities since 1960. Business credit grew quite rapidly in the mid-1980s and was followed by a substantial credit slowdown at the end of the decade. When measured in
nominal terms, however, the growth in the mid-1980s does not appear all that unusual
when compared to the late 1960s, mid-1970s, and late 1970s.
This comparison is not particularly meaningful, though, because inflation (as well as
the growth in economic activity) was greater in those earlier periods. In order to make
the measured growth rates more comparable over time, we have also included in the top
panel of Chart 1 the real rate of credit growth. The enormity of the run-up in debt in the
1980s and sharpness of subsequent slowdown becomes much more evident when measured on an inflation-adjusted basis. The sharp rise in leverage and subsequent slowdown is even more apparent in the lower panel of Chart 1 where we have graphed the
ratio of total business liabilities to business sector GDP. Here we see that (1) a rising
trend in leverage is evident throughout the sample period but it accelerated sharply in
the mid-1980s, and (2) the recent slowdown in credit growth has led to a decline in the
debt-to-GDP ratio. Despite the recent sharp deceleration in credit growth, the ratio of
credit to GDP has fallen only modestly because the growth of GDP also slowed at the
same time.
As in previous periods of credit stringency, data that has been aggregated across
many funding sources are insufficient to distinguish whether the credit slowdown was
primarily the result of supply or demand factors. Researchers have, however, been able
to identify supply-side credit crunches in the past by observing occasions in which frustrated borrowers shifted away from bank credit to nonbank funding sources.5 Chart 2
suggests that frustrated borrowers at banks were not able to shift to nonbank sources of
credit during the recent credit slowdown. On this chart, we have plotted the levels of
5
See, for example, (1) Anil Kashyap, Jeremy Stein and David Wilcox, "Monetary Policy and Credit Conditions: Evidence from the Composition of External Finance," American Economic Review, March 1993;
(2) Mark Gertler and Simon Gilchrist, "Monetary Policy, Business Cycles and the Behavior of Small Manufacturing Firms," NBER Working Paper #3892, 1992; and (3) Stephen Oliner and Glenn Rudebusch,
"The Transmission of Monetary Policy to Small and Large Firms," mimeo, Federal Reserve Board, June,
1992.
bank loans, commercial paper, finance company loans, and corporate bonds—all benchmarked to 100 at July 1990. Around that date, the rapid growth of finance company
loans and commercial paper abruptly halted and the rate of bond growth slowed slightly.
Table 1 presents data for selected liabilities (excluding commercial mortgages and
deferred taxes) of the nonflnancial business sector over various periods. A brief description of these sources of business credit is given in the Appendix. The table shows that
the growth in total liabilities fell from about 11 percent average annual growth rate between year-end 1983 and year-end 1988 to 4 percent annual growth between year-end
1988 and the year-end-1992. Over this latter period, bank loans grew much more slowly
Chart 1: Total Nonf inancial Business Sector Liabilities
Percent
20
Source: Board of Governors, Flow of Funds and Department of Commerce, National
Income and Product Accounts.
Note: The charts display credit to nonfarm, nonfinancial business.
175
Causes and Consequences
(virtually no growth) than did commercial paper (6 percent annual growth), bonds (6
percent annual growth), and finance company loans (5 percent annual growth).6
This simple comparison creates an impression that the credit crunch was primarily a
banking industry phenomenon. The correct interpretation, however, is likely more complicated. While bank credit growth fell from about 8 percent per year to about nil, the
growth rates of most of the other components of business credit, excepting trade debt
and corporate bonds, fell by at least as many percentage points, albeit from higher initial
rates. The analysis presented below, which takes into consideration the secularly declining share of C&I loans in total business credit, suggests that credit restraint was common
to many business financing sources over the last few years.
Graphical Comparisons of the Current and Previous Slowdowns
Chart 3 examines the real growth rates of bank C&I loans and nonbank credit—consisting of corporate bonds, commercial paper, finance company loans, and other loans
(bankers acceptances, loans from savings and loan associations, and foreign loans)7 —
from eight quarters before to eight quarters after the onset of the recession. One of the
lines on each chart represents real credit growth over the recent period (1988 to 1992)
and the other represents an average of credit growth over four periods in which credit
slowed abruptly.8 Here we see that (1) the slowdown in bank loans was more gradual
and began earlier in the current episode than in past credit crunches, and (2) nonbank
6
Bank loans also grew slightly less than commercial mortgage lending but the drop in mortgage lending
growth was significantly greater than in any other component of lending.
7
The financial data for Charts 3-6 arc for the nonfarm, nonfinancial business sector from the Flow of Funds,
Board of Governors of the Federal Reserve System. The data on the gross domestic product and the price
deflator arc from the National Income Accounts published by the Department of Commerce.
8
The four earlier episodes arc centered around the following dates 1965-1V, 1969-1V, 1973-1V, and 1981III. The 1965 period covers the 1966 credit crunch (bank lending growth on a four quarter basis began to
Chart 2: Selected Liabilities of the Nonfinancial Business Sector
176
July 1990=100
130
credit growth slowed continuously through the recent period; whereas, in past episodes,
nonbank credit has tended to increase while bank lending was slowing.
In past periods of credit restraint, nonbank funding growth lagged bank loan growth
fairly consistently—(1) trailing the surge in bank lending at the cycle peak, (2) growing
more quickly than bank loans as the crunch began, and (3) eventually flattening out or
slowing about a year later. Previous researchers have interpreted this pattern as evidence that, in these episodes, supply-side shocks to bank credit have led frustrated borrowers to substitute to nonbank sources of funds. In contrast, over the last few years,
this substitution has not been evident. Of course, the bank loan growth level, as opposed
to its rate of decline, is much lower than nonbank loan growth. This differential growth
rate, however, opened up long before the recession and the difference has actually narrowed during the recent period of slow credit growth.
Footnote 8 continued
decline in 1965-1V) while the other central dates correspond to business cycle peaks. The 1980 credit
crunch was not included because, being so brief and occurring so shortly before the next crunch, its data
overlap with data from the 1981-82 recession. For the current recession, we centered the data around the
third quarter of 1990, the NBER reference date for the business cycle peak. A reasonable alternative date
for a recent turning point for the current credit cycle might be earlier, perhaps somewhere in mid-1989
when interest rates peaked and bank credit and economic activity had already started to slow.
Table 1: Business Credit Growth, Various Periods
Growth Rates Measured at Annual Rates, Dollar Figures in Billions
1992-1V
Level $ Bil.
1960-69
1970-79
1980-82
1983-88
1989-92
Bank loans
$661
11%
10%
12%
8%
0%
Finance company loans
297
13%
15%
6%
16%
5%
Commercial paper
108
n.a.
17%
15%
15%
6%
Trade debt
811
8%
8%
8%
8%
4%
Corporate bonds
1231
7%
10%
9%
12%
6%
$3307
9%
10%
10%
11%
4%
$6082
7%
10%
7%
8%
5%
122
3%
7%
8%
4%
4%
$1022
10%
11%
6%
14%
1%
Credit source
Total liabilities8
Activity measures
GDP (current dollars)
GDP deflator (1987=100)b
Memo item
Commercial and
multifamily mortgages
Source: Board of Governors, Flow of Funds, and Department of Commerce, National Income
and Product Accounts.
a
- Category includes items listed above plus other loans that are not shown separately: loans
from thrifts, loans from foreigners, and bankers' acceptances.
b
- Implicit deflator for GDP.
177
Causes and Consequences
Chart 4 shows that economic activity9 and inventory accumulation, the primary determinants of short-term credit demand, were weaker before the recent recession than
during periods before previous recessions. We next consider whether firms with limited
credit access may have attempted to finance additional inventory accumulation by increasing their trade debt, perhaps by stretching out their payables. The aggregate (gross)
trade debt of the nonfinancial business sector is shown in Chart 4. Here we see that trade
debt slowed more gradually and earlier than in past cycles, consistent with the economic
Activity is measured by the real gross domestic product for the nonfinancial business sector.
Chart 3: Business Credit Growth
178
Note: The current recession is "centered" around 1990-111. The other cycles are centered
around 1965-IV, 1969-IV, 1973-1V, and 1981-111.
fundamentals. Hence, there is no clear evidence here that firms substituted away from
credit markets to interfirm trade debt.
We turn now to a more detailed discussion of the components of nonbank credit.
Nonbank funding did not surge in the beginning of this credit crunch as it had in past
credit crunches (Chart 5). This observation holds particularly strongly for the commercial paper market. Unlike past credit crunches, in which large firms with access to markets clearly shifted away from bank loans to commercial paper, no such shift is evident
in the current episode. Moreover, the amount of recent substitution away from bank
loans to finance company loans or other funding sources (bankers' acceptances, loans
by S&Ls, and foreign loans) seems small compared to past cycles.
So far the discussion has been limited to short- and intermediate-term credit. Some
borrowers may have shifted from intermediated credit and commercial paper into bond
financing during periods of credit restraint. As shown in Chart 6, real bond growth is
normally strongly countercyclical,10 and the current episode appears to be conforming
10
The cyclical pattern of bond issuance is analyzed by Benjamin Friedman in "Substitution and Expectations
Effects on Long-Tcrm Borrowing Behavior and Long-Term Interest Rates," Journal of Money, Credit, and
Banking, May, 1979. Interest rates cycles have coincided with business cycles and hence borrowers are
likely reluctant to issue long-term bonds at the beginnings of recessions. Moreover, investment spending
plans are typically scaled back at business cycle peaks. In the current recession, the decline in bond issuance pattern is probably better explained by the vicissitudes of the junk bond market, discussed below.
Chart 4: Real Business Activity and Trade Debt
179
Causes and Consequences
with previous patterns—although the decline before the recession was deeper and the
rebound after was weaker than in past episodes.
The current credit cycle is complicated by the unusual behavior of net equity issuance since the early 1980s. In the latter half of the 1980s, net equity issuance was sharply negative as a result of M&A/LBO activities and stock repurchases. Consequently,
total net financing from long-term capital markets (netting out the equity retired and
bond growth) was not as strong as bond growth. When long-term funding growth is
measured by bond growth adjusted for net equity issuance, funding growth appears
slower than bond growth before the recession but faster afterwards. The recent strength
in long-term securities issuance along with the relative weakness in commercial paper
issuance may be explained, in part, by a shift toward longer-term financing by firms with
access to organized credit markets in reaction to currently attractive long-term borrowing rates.
Regression Analysis
In this section, we examine in a more formal manner whether the recent behavior of
bank and nonbank credit has been unusual in light of their traditional comovements with
Chart 5: Short and Intermediate Term Business Credit from Nonbanks
Note: The current recession is "centered" around 1993-111. The other cycles are centered around 1965-IV,
1969-1V, 1973-1V, and 1981-111.
180
other macroeconomic variables. If the credit crunch were primarily reflective of problems facing the banking system, we would expect bank and nonbank credit to exhibit
substantially different patterns of deviation from their historical relationships with other
variables. In particular, borrowers would have substituted away from bank loans to nonbank credit—those with market access would have shifted to the bond and commercial
paper markets and others would have turned to finance companies. Our regression analysis indicates that, even after accounting for the general weakness in economic activity,
both bank and nonbank credit grew surprisingly slowly over the last three years. When
measured as simply the number of percentage points by which the growth predicted by
econometric equations exceed actual growth, the shortfall in bank lending was greater
than the weakness in nonbank credit. If, however, the data is normalized to account for
the historically greater volatility of bank credit relative to nonbank credit, the relative
deviations between realized and predicted growth for the two credit aggregates appear
more similar.
The analysis below compares bank loans to a nonbank credit aggregate consisting of
nonfinancial corporate bonds, finance company loans and leases, and nonfinancial commercial paper. The data for these four sources of business finance are available on a consistent, seasonally adjusted basis from the Flow of Funds Release prepared by the Board
of Governors of the Federal Reserve System.11 We have chosen to study nonbank credit
on an aggregated basis, rather than by individual components, because (1) the relative
importance of different sources of finance has shifted dramatically over time and (2)
1
' We have not included certain components of nonbank business credit—such as thrift lending, government
loans, and foreign loans—either because they are measured highly imperfectly at a quarterly frequency or
because they are so small as to be negligible in the aggregate. We considered including trade credit in the
nonbank credit aggregate but found that trade credit was extraordinarily correlated with GDP and inventories and suspect that its independent variations are dominated by measurement errors.
Chart 6: Nonfinancial Corporate Bonds
Note: The current recession is "centered" around 1990-111. The other cycles are centered
around 1965-IV, 1969-IV, 1973-IV, and 1981-111.
181
Causes and Consequences
182
considerable substitution between short- and long-term financing occurs over time for
reasons other than general credit availability.12
Although our econometric specifications include only lagged macroeconomic variables as determinants of credit growth, causality likely runs both ways between credit
growth and economic activity. Our regressions should therefore be thought of as semireduced-form relationships designed to distinguish normal and unusual comovements in
the data but incapable of independently identifying supply and demand shocks. 13
Our basic regression models are given in Table 2. Here we relate the annualized
quarterly percentage growth rates of bank loans and nonbank credit to growth rates of
the stock of inventories, the stock of producer's durable equipment, and GDP of the nonfinancial business sector. Each regression includes two lags of each explanatory variable plus two lags of the dependent variable. (Estimates with four lags of each variable
produced very similar results.) The estimation period, 1960-IV to 1992-IV, is determined by the first date at which business GDP is available and the required lags.
In light of the fact that the equations are specified in growth rates and that flow of
funds data contain a lot of noise, the adjusted R2, 0.35 and 0.52, for the bank and nonbank equations, respectively, are quite reasonable. Moreover, various robustness checks
(e.g., splitting the sample, adding and deleting variables) revealed that the patterns of
the regression errors were largely impervious to reasonable alternative specifications.14
While the estimated coefficients are consistent with various stories one might propose,
one should not place too much weight on specific coefficients because of the high degree
of multicollinearity in the explanatory variables and the mixture of the supply and demand effects in the data. The residual standard error of the bank loan equation is more
than twice as large as that of the nonbank equation, reflecting the weaker fit of the former
equation and the greater variability of bank loans more generally. We therefore should
not be surprised to find larger prediction errors for the bank equation than for the nonbank equation in any given subsample period, including the recent credit crunch.
The first line in Table 3, presents a very crude quantitative measure of unusual components of the credit slowdown. Here, we simply appended a dummy variable for the
last twelve quarters (1990-1 to 1992-IV) to our basic regressions. The estimated coefficients imply that banks loans grew slower than expected by 5.5 percent on an annualized
basis; whereas, nonbank credit was 2.6 percent below expectations. We then inserted
dummy variables for the last five recessionary periods in the specification, treating the
mini-recession of 1980 as part of the 1981-1982 recession and including the slowdown
(and credit crunch) of 1966 as if it were a recession. Here we see roughly the same pattern for the three quarters of the recent recession as for the last twelve quarters reported
before. With regard to the other recessionary periods, only the 1966 experience is clearly identifiable as a period of slower than expected credit growth.
12
Moreover, preliminary regression analysis on the disaggregated components revealed an unusual amount
of instability between the credit components and macroeconomic determinants. The standard errors on
individual coefficient estimates and for the regressions as a whole were unusually large.
13
Our approach and purpose is very similar to the time series analysis conducted by Cara Lown and John
Wenninger in their paper in this volume. For a formal attempt to identify the feedback to reduced credit
availability on credit demand, see Patricia Mosser's paper in this volume.
14
We experimented with including interest rates in the specification. However, we did not include interest
rates in our final specifications because (1) we had trouble identifying the appropriate nonbank rate and (2)
credit growth has historically been positively correlated with rising rates, leading to a positive coefficient
estimate (presumably reflecting movements along a supply curve for credit in response to shifting
demand) even though the basic specification appears closer to a demand than a supply equation.
Table 2: Bank and Nonbank Credit Growth Equations
Dependent Variables
Explanatory Variables
Bank Loans
Nonbank Credit
Dependent variable M
0.25
(2.65)
0.26
(2.84)
Dependent variable ,.2
0.15
(1.57)
0.18
(2.00)
Inventories
-0.13
(0.43)
0.05
(0.37)
Inventories t_2
0.48
(1.34)
0.20
(1.31)
Investment
M
7.64
(1.97)
1.29
(0.80)
Investment t.2
-6.13
(1.77)
-0.45
(0.30)
Business GDP M
0.33
(0.42)
0.71
(2.11)
Business GDP t.2
0.55
(0.94)
0.21
(0.84)
H
2
0.35
0.52
Standard error
6.57
2.76
Durbin-Watson
2.06
2.07
Adjusted R
Sources: Commerce Department and Flow of Funds.
Notes: Estimation period: 1960-1V to 1992-1V. Absolute t-statistics reported in parentheses.
Regressions include a constant term, not reported above. All data are measured in nominal
annualized percentage growth rates. Inventories = net growth in the stock, exclusive of price
revaluations. Investment = gross investment in producer's durable equipment divided by the
capital stock. Nonbank Credit = bonds + finance company loans + nonfinancial commercial
paper.
Table 3: Coefficients on Dummy Variables for Credit Growth Equations
Period of Dummy Variable
Bank Loans
Nonbank Credit
The recent experience
1990-1 to 1992-1V
-5.53
(2.38)
-2.58
(2.58)
1966-111 to 1966-1V
-8.78
(1.79)
-4.47
(2.15)
1969-1V to 1970-1V
-3.74
(1.20)
-0.99
(0.75)
1973-1V to 1975-1
-0.51
(0.17)
0.26
(0.02)
1980-1 to 1982-1V
5.33
(2.20)
1.15
(1.16)
1990-111 to 1991-1
•4.18
(1.09)
-2.90
(1.77)
Five recessionary periods
Notes: Absolute t-statistics are reported in parentheses. Regressions are estimated over
1960-1V to 1992-1V. The regression specifications and data definitions are those in Table 2.
183
Causes and Consequences
We next examine the prediction errors for the last twelve quarters in detail. We reestimated the equations using data through year-end 1989 only and obtained the equations' predictions for the next twelve quarters. The results are presented in Table 4.
Quite remarkably, the prediction errors are negative for every quarter for both the bank
and nonbank specifications.15 The average quarterly errors on a annualized basis are
5.7 percent the bank equation and 2.6 percent for the nonbank equation. (The close similarity to the dummy variable approach is not surprising.) The F-statistics based on the
root-mean squared errors of the forecasts suggest that we cannot reject the stability of
the nonbank and, particularly, the bank equations. However, these tests are highly misleading because they look only at the size of the errors and not their direction. As dis15
We calculated static rather than dynamic forecasts; i.e., we inserted observed rather than forecasted values
of the lagged dependent variables into the projections. Given the positive coefficients on the lagged
dependent variables in both equations, this choice biased us against finding a large credit slowdown and
against finding consecutive quarters in which the forecast errors were negative.
Table 4: Bank and Nonbank Credit Growth Equations Recent Forecast Errors
184
Nonbank Credit
Bank Loans
Actual
Predicted
Error
Actual
Predicted
Error
1990-1
1.1
4.4
-3.3
4.7
7.1
-2.3
1990-11
3.4
6.3
-2.9
6.1
7.3
-1.1
1990-111
-0.9
4.1
-4.9
3.6
6.8
-3.2
1990-1V
0.3
7.0
-6.7
1.5
6.5
-5.1
1990-1
-1.6
3.5
-5.2
2.0
4.1
-2.1
1991-11
-7.9
-0.3
•7.6
1.5
2.9
-1.5
1991-111
-3.2
0.3
-3.5
1.8
3.8
-2.0
1991-IV
-9.0
2.1
-11.1
-1.6
4.3
-5.8
1992-1
-2.1
1.0
-3.1
3.2
3.6
-0.4
1992-11
-4.9
2.4
-7.4
1.6
4.5
-2.9
1992-111
-1.0
5.0
-6.0
3.1
5.1
-2.0
1992-1V
-1.7
4.8
-6.5
3.2
5.6
-2.4
Average error of the
forecast
-5.67
-2.56
Standard error of the
regression
6.72
2.78
Root mean squared
error of the forecast
6.12
2.95
F-test significance
0.62
0.35
Notes: Absolute t-statistics are reported in parentheses. Regressions are estimated over
1960-1V to 1989-1V and forecast dynamically from 1990-1 to 1992-1V. The regression specifications and data definitions are those in Table 2.
cussed above, the dummy variable approach shows that the average error is statistically
different from zero. Moreover, the probability of getting twelve consecutive negative
forecast errors from a stable regression equation is infinitesimal—roughly 1/2 raised to
the 12th power. Clearly, the recent credit slowdown is not adequately explained by these
regressions.
In order to see which category of lending declined the earliest and by the largest
amount, we plot in Chart 7 both the in-sample regression errors and the out-of-sample
forecast errors for both bank and nonbank regression errors. The data are presented on
a four-quarter-moving-average basis and have been normalized by dividing by each
equation's standard error so that the vertical axis measures the number of standard deviations that the errors stray from the mean error, zero. We see that in most periods, the
bank and nonbank errors move together. In the mid-1980s, however, nonbank credit
growth was unusually rapidly relative to bank loans. Starting from year-end 1986, the
bank and nonbank credit prediction errors begin to decline in almost lock-step fashion
until they reach their troughs, in early 1990 for nonbank credit and early 1991 for bank
loans. These patterns argue strongly against the view that the credit crunch has been primarily a banking phenomenon. In fact, the standardized prediction errors of the nonbank equation have fallen more since 1986 than the bank equation errors, once we take
into account the historical greater variability of bank lending.
Taken together, the evidence presented in this section suggests that the credit crunch
was spread broadly across sources of business credit. Since economic activity slowed
at the same time, one might conclude that the credit slowdown was primarily a demand
phenomenon reflecting the slowdown in economic activity. However, the regression
analysis indicates that it is difficult to explain the decline in credit growth on the basis
of simple historical correlations with macroeconomic variables alone. The impression-
Chart 7: Banks and Nonbank Credit Growth Prediction Errors
Four-Quarter Moving Average
Standard deviations
1.5
1960 62
64 66
70
72 74
76
78
80
82
84
86
88
90 92
Note: Data were normalized by dividing both in-sample residuals and out-of-sample forecast errors by the estimated standard errors of the regressions from which they were
derived.
185
Causes and Consequences
istic evidence reported in the next section, which describes the circumstances of various
nonbank intermediaries and money and capital markets during this period, argues that
supply factors likely played a significant role in the credit slowdown.
Factors Affecting Nonbank Credit Availability
Simple quantitative analysis of aggregative data ignores considerable available information concerning the recent credit crunch. The following section reviews the recent
stresses experienced by nonbank financial intermediaries and the fragility visible in
parts of the organized credit markets. This evidence suggests that credit availability
from nonbank funding sources has been constraining for some borrowers, particularly
firms that are small or rated below-investment grade. The asset-backed securities market, however, did help relieve some of the credit restraint pressures by easing balance
sheet constraints at financial intermediaries. The following subsections discuss these
developments at finance companies, life insurance companies, and in the commercial
paper market, organized capital markets, and asset-backed securities markets.
Finance Companies
186
Financial problems among finance companies began at about the same time as those of
the banks of New England.16 Beginning in late 1989, many finance companies were
downgraded by the credit rating agencies either because the financial condition of their
parent companies had deteriorated or because they themselves had suffered major losses
in their commercial lending businesses. These credit rating downgrades may have had
a significant effect on lending because most finance companies raise the majority of
their funds in short-term public credit markets.17
Following poor sales in the late 1980s, the major U.S. auto companies were repeatedly downgraded by the rating agencies and the ratings of their captive finance subsidiaries were similarly affected (see Chart 8A). In June of 1989, under threat of a further
downgrade from already low ratings, Chrysler Financial Corporation gradually withdrew from the commercial paper market and began borrowing under more expensive
bank facilities.18 In addition, Chrysler stepped up its auto loan securitization program
in order to reduce significantly the amount of loans on its books. The short-term credit
obligations of GMAC appeared at times to be on the verge of being downgraded so that
it too might have trouble funding itself in the commercial paper market.19 The finance
company commercial paper programs are generally backed up by credit lines at commercial banks. In response to their funding problems in the commercial paper market,
finance companies increased their borrowings from banks (see Chart 8B), reversing a
16
The problems in the banking system became increasingly apparent in the Fall of 1989 and throughout
1990, with the Fall of 1990 being the low point if one uses the stock market as a guide.
17
In a pooled time-scries, cross-sectional study, Rcmolona and Wulfckuhlcr show that credit ratings arc the
single most important determinant of finance company asset growth. See Eli Remolona and Kurt
Wulfckuhler, "Finance Companies, Bank Competition, and Niche Markets," Quarterly Review, Federal
Reserve Bank of New York, Summer, 1992.
18
At year-end 1989, Chrysler Corporation had $10 billion in commercial paper outstanding and no bank
loans. At year-end 1990, it had $1 billion in commercial paper outstanding and over $6 billion in bank
loans.
19
Weak sales by McDonnell Douglas and subsequent downgrades also drove its finance company from the
CP market around the same time as Chrysler. GMAC's commercial paper rating was ultimately downgraded to P2 by Moody's in late 1992 and downgraded to A2 by Standard & Poors in early 1993.
Chart 8A: Senior Debt Ratings of Financial Companies
Rating
Source: Moody's Investor Service.
Note: Sample consist of 22 life insurance companies, 20 banks, and 18 finance companies.
Institutions that defaulted during period are not included.
Chart 8B: Finance Company Liabilities: Ratio of Banks Loans to Commercial Paper
187
Causes and Consequences
long-term trend decline. Anecdotal evidence, however, suggests that the cost of backup
lines has risen over the last few years, so that credit stringency at banks has had an adverse feedback on credit availability at finance companies. 20
Diversified finance companies—those finance companies that make significant
amounts of business loans (as opposed to firms that specialize in consumer finance)—
also suffered numerous credit rating downgrades during the recession.21 In the aggregate, finance companies were forced to put up unprecedented, large amounts of loan
loss provisions (see Chart 8C) and subsequently experienced high charge-off rates (see
Table 5).
Despite increased loss rates and rating downgrades, many finance companies were
able to maintain reasonable profitability in 1990 and 1991 because of the dramatic decline in short-term interest rates and, hence, funding costs. Profitability remained particularly strong for finance companies that specialize in consumer lending (i.e., credit
card loans, personal loans, sales finance loans, and home equity loans) because their loss
experiences did not rise too severely in the recession.
20
Independent mortgage bankers without access to the commercial paper market have traditionally relied on
banks to fund inventories of mortgages until the inventory can be sold. Banks completely withdrew
financing from mortgage bankers in late 1990 and 1991. The aggregate supply of mortgage credit was not
significantly affected, however, because alternative means of financing were successfully arranged. These
views were expressed by Lyle Gramley, Chief Economist of the Mortgage Bankers Association, at a colloquium on the credit crunch held at the Federal Reserve Bank of New York on February 12, 1993.
21
The problems of Westinghouse Credit, made public in the beginning of 1990, were the most extreme: over
a twelve month period, the company wrote off about $2.5 billion in commercial real estate loans, about 25
percent of its entire lending portfolio. (At year-end 1993, Westinghouse announced it was liquidating its
finance subsidiary and expected $2.5 billion in additional, related losses.) The Westinghouse Credit shock
did not, however, shatter the market for finance company securities because (1) Westinghouse Corporation
stood behind its troubled subsidiary and (2) no other major finance company had that kind of exposure to
commercial real estate although a few took significant real estate-related hits to earnings.
Chart 8C: Loan Loss Reserves at Finance Companies
188
From 1984 to 1988, total business credit extended by finance companies grew at an
extraordinarily rapid pace—between 15 percent to 20 percent per year. Over the next
two years, finance credit growth slowed considerably but continued to be stronger than
bank loan growth. The source of this strong credit growth, particularly in recent years,
has been primarily growth in leasing, which accounts for over one-third of all finance
company business receivables outstanding.22 The sluggish growth rate of the nonleasing component of finance company lending from late 1989 onward has been quite striking (Chart 9). The continued strength of leasing receivables growth, suggests that
finance companies were more willing to make collateralized loans than unsecured loans
during the recession. The decline in the nonleasing component of finance company
credit presumably reflects not only the funding problems and lower credit ratings of the
finance companies, themselves, but also the decline in demand and the tightening of
credit standards that normally occurs in a recession.
The continuing presence of finance company assets "on the block and up for sale"
may have reduced the incentive and capital support for asset growth through loan origination at healthy firms. The finance company industry has been consolidating in recent
years as healthy firms, particularly GE Capital and Associates Corporation, have acquired the assets of weaker competitors. Finance companies that face high funding
22
The tax and accounting benefits of leasing are discussed in the paper by Remolona and Wulfekuhler that
was cited above. The ability of borrowers to shift from bank C&I loans to leasing is quite limited in the
short run. C&I loans are often for general corporate purposes and arc not intended to finance particular
pieces of equipment that could alternatively be leased.
Table 5: Performance Measures for Finance Companies
In Percent
Average 1987-89
1990
1991
Auto finance firms
8.0
0.9
-3.2
Consumer firms
13.9
14.4
11.3
Diversified firms
22.3
14.0
7.0
Auto finance firms
1.2
1.1
1.3
Consumer firms
2.1
2.0
1.9
Diversified firms
1.8
1.0
0.5
Auto finance firms
0.9
0.9
1.1
Consumer firms
1.7
2.2
2.6
Diversified firms
1.0
3.0
3.5
Growth in assets
Return on assets
Net chargeoffs (percent of receivables)
Note: Table is based on estimates derived by Mitchell Post, Board of Governors, from annual
10K filings for largest twenty-five finance companies, comprising about three-quarters of the
industry's assets.
189
Causes and Consequences
costs in the capital markets cannot compete and must either shrink assets or seek a stronger acquirer. Many of the finance companies whose assets were acquired by others were
in fact healthy though hobbled in the capital markets by the troubles of their parents.
U.S. bank holding companies, in particular, have been divesting themselves of their finance company subsidiaries in order to raise capital. Table 6 shows that, since 1988,
finance company loans at subsidiaries of U.S. bank holding companies have been
shrinking at a rapid rate, in fact, much more so than total loans at these bank holding
companies on a consolidated basis.
Life Insurance Companies
In past credit crunches, life insurance companies had to reduce their investments, particularly affecting the private placements market, because they, like depository institutions, experienced disintermediation as interest rates rose.23 At those times, traditional
life insurance products could not compete with variable rate savings instruments and,
hence, insurers lost funds at the peak of each interest rate cycle. During the present cycle, however, life insurance company liabilities did not have to shrink because interest
rates did not rise sharply and insurers now rely more on variable rate products for their
funding.
During the current recession, some venerable firms in the life insurance industry suffered credit downgrades and faced unusual scrutiny by critics in Congress and the press
for poor asset quality, insufficient capital, and inadequate regulatory supervision. In
Past episodes of disintermediation at life insurance companies and the impact on the private placement
market is discussed in Patrick Corcoran, "A Cyclical Perspective on Fixed-Income Private Placements,"
mimeo, Prudential Insurance, September, 1991. It is also worth noting that funding needs often rose at life
insurance companies at interest rate cycle peaks due to increased demand by policyholdcrs for (low interest) policy loans.
Chart 9: Components of Finance Company Business Credit
190
July 1990=100
January of 1990, First Executive Corporation announced a large write-down of assets
because of losses on its junk bond portfolio.24 In October of the same year, The Travelers announced substantial losses on its commercial real estate portfolio. Contagion effects emanating from these firms and from problems in the banking sector caused
investors to question the soundness of many insurers, particularly after Equitable announced large losses, and First Executive, First Capital, and Mutual Benefit failed.
Most of the industry's problems stemmed from commercial real estate lending (see
Chart 10A), 25 poorly performing junk bond portfolios, and overly generous rates of return that had been promised to guaranteed investment contract investors in the mid1980s.
The problems faced by insurers were similar to those faced by banks. Commercial
real estate and highly-leveraged-transactions-related lending losses precipitated the
troubles, but these problems were compounded by severely negative press coverage and
public scrutiny following some prominent failures. The plunge in life insurance company stock prices occurred at the same time as the fall in bank stocks, although insurance
stocks did not fall as far on average (see Chart 10B). The credit ratings of insurers also
fell, but again by less than the ratings of other financial intermediaries (see Chart IOC).
Insurers became preoccupied with maintaining liquidity and preserving their reputations for financial security, hoping to avoid the fate of Mutual Benefit Life, which col24
First Executive's problems were first evident in 1987 when regulators in California and New York required
that the holding company downstream equity to the two insurance subsidiaries in those states.
25
The severity of the recent problems in commercial real estate is most evident in the unprecedented rate of
foreclosures and not as evident in the rate of mortgage delinquencies. Real estate losses are primarily recognized when a foreclosure takes place, and a rise in foreclosures actually reduces the reported delinquency rate.
Table 6: Loans at Bank Holding Companies and Their Finance Subsidiaries
Year-End, in Billions of Dollars
1988
1989
1990
1991
1,533
1,658
1,703
1,619
C&l loans and leases
498
524
507
455
Real estate loans
542
615
671
685
Consumer loans
287
300
302
292
Other loans
95
102
97
84
54.4
43.2
35.3
26.4
C&l loans and leases
n.a.
n.a.
17.2
12.3
Consumer loans
n.a.
n.a.
6.6
5.4
Real estate loans
n.a.
n.a.
10.1
7.4
Other loans
n.a.
n.a.
1.4
1.1
U.S. bank holding companies
Total loans and leases
Finance company subsidiaries
Total loans and leases
Source: FR Y-9C and FR Y-11AS filings.
191
Causes and Consequences
lapsed suddenly in the middle of 1991 as a result of a run on its liabilities by pension
investors.
As a result of the problems of Executive life and other insurers, the National Association of Insurance Commissioners ("NAIC") adopted rules in mid-1990 that required
greater disclosure and reserves against below-investment grade bonds. In addition, the
new rules required that insurers classify (based on internal ratings) private placements
as if they were public firms for reserves and capital purposes. These developments not
only reduced the willingness of insurance companies to invest in below-investment
grade bonds (see Table 7), they induced a shift toward low risk assets more generally.
These developments appear to have restrained credit availability in the market for
privately placed bonds, an important source of finance for below-investment-grade or
unrated middle market firms.26 The ratio of private placements to public offerings of
corporate bonds has plummeted over the last three years (Chart 11). While a relative
decline in private placement issuance is not unusual going into a recession, the depth
and persistence of the recent decline is surprising.27 The new NAIC rules and additional
scrutiny of insurer portfolios by rating agencies and investors did apparently affect the
26
This section borrows heavily from the analysis contained in a paper by Mark Carey, Stephen Prowse, John
Rea and Gregory Udell, " Recent Developments in the Market for Privately Placed Debt," Federal
Reserve Bulletin, February, 1993 and a paper by Patrick Corcoran, "The Credit Slowdown of 1989-1991:
The Role of Demand," published in Credit Markets in Transition, Federal Reserve Bank of Chicago, 1992.
27
The share of privately placed bonds in total bond issuance has historically followed a clear cyclical pattern. The private market's share tends to fall in the latter half of expansions, to remain low in recessions,
and to pick up early in the recovery. The common explanation for this pattern prior to the 1980s is that life
insurance companies, which are the primary investors in the private market, experienced disintermediation
when interest rates rose. Today, however, insurers' liabilities arc more interest sensitive and, thus, investments in private placements tend to decline late in the cycle as borrowing demand decreases. (These
developments are discussed in Patrick Corcoran, "A Cyclical Perspective on Fixed-Income Private Placements," op. cit.)
Chart 10A: Problem Commercial Mortgages at Life Insurance Companies
192
Percent
5
Chart 10B: NASDAQ Stock Indexes
Chart 10C: Senior Debt Ratings of Financial Institutions
Rating
Aa1
Source: Moody's Investor Service.
Note: Sample consists of twenty-two life insurance companies, twenty banks, and eighteen
finance companies.
Institutions that defaulted during period are not included.
193
Causes and Consequences
willingness of insurance companies to purchase below-investment-grade private placements. The share of these purchases as a proportion of life insurance companies' private
placement investments dropped off sharply in 1990 and has not recovered (see Chart
11). Moreover, the spread between private and public below-investment-grade bonds
widened at the same time.28
The Commercial Paper Market
Unlike earlier credit crunches, on this occasion nonfinancial firms did not increase their
rate of issuance of commercial paper (see Charts 2 and 5). Chart 12, which depicts the
spread of commercial paper rates over comparable Treasuries, suggests that during the
credit slowdown period, the market was quite receptive to new issuance. While this interpretation may be correct for the highest quality issuers, the CP market has expanded
in recent years to lower quality issuers and for them the recent story is quite different.
Perceived credit risk in the commercial paper market grew in the late 1980s because
of a series of defaults and an increase in the number of credit rating downgrades relative
to upgrades in 1988 and 1989 (See Chart 12). Prior to June of 1989, there were only
two defaults in the history of the U.S. commercial paper market, Penn Central in 1970
and Manville in 1982. However, ten issuers defaulted during the year prior to the recession and five additional issuers have subsequently defaulted. Although many of these
defaulting issuers were small, obscure and unrated, the confidence that investors, particularly mutual fund investors, previously had in the safety of commercial paper was rat28
Although data are not readily available, the papers cited in the previous footnote state that dealers in private placements at investment banks reported a sharp widening of spreads between privately and publicly
placed below-investment-grade securities during 1990 which has continued through 1992.
Table 7: Below-lnvestment-Grade Bond Holdings of
Twenty Large U.S. Insurance Companies
Thousands of dollars
194
(A)
(B)
(C)
(D)
(E)
Total Bond
Holdings
Rated "B" or
Below
Holdings
Rated "BB"
Holdings
(B)/(A)
(C)/(A)
1987
$211,637
$17,545
n.a.
8.3%
n.a.
1988
255,089
17,810
n.a.
7.0
n.a.
1989
281,881
19,604
n.a.
7.0
n.a.
1990
303,548
17,504
$15,659
5.8
5.2%
1991
334,965
16,502
13,962
4.9
4.2
1992
350,186
15,523
13,330
4.4
3.8
Year
Source: Conning and Company; Federal Reserve Bank of New York estimates.
Notes: Data for TIAA-CREF not available. Table does not include some companies with large
holdings of low-rated bonds, but relatively small total bond holdings, First Capital and
Executive Life.
Sample: The top twenty life insurance companies in terms of corporate bond holdings.
tied. In at least one default, two money market funds were faced with the possibility
that their share values would fall below one dollar, prompting parents of the funds to
buy defaulted commercial paper at a loss in order to avoid "breaking the buck."
In response, mutual funds began to shun all commercial paper except that with the
highest ratings in mid-1990. The industry was concerned that some mutual fund might
end-up breaking the buck because it had over-invested in risky commercial paper. Such
an event might trigger a "run" on the entire industry. In order to protect small investors
and the reputation of the money market fund industry, the SEC adopted rules in July
Chart 11: Indicators of Demand for Privately Placed Bonds
Percent
Sources: SEC, Investment Dealers Digest, FRB Capital Markets, American Council of Life
Insurance.
195
Causes and Consequences
1990 that imposed strict limits on the amount the so-called "second tier" paper that
funds could hold.29 At the time of the announcement, the Investment Company Institute, a trade group that represents the mutual funds industry, made a statement to the
press that the new rules were perhaps not strong enough.30
Chart 13 presents the total amount of low quality commercial paper held by mutual
29
These rules were formally announced on July 25, 1990. Sec "Revisions to Rules Regulating Money Market Funds: Proposed Amendments to Rules and Forms," (Release Nos. 33-6870; IC-17589, S7-13-90)
SEC Docket., vol. 46, pp 1247-61.
3 0 ••;
Money Market Funds Shedding Lower-Grade Paper," Wall Street Journal, October 22, 1990, p. Cl.
Chart 12: Short-Term Corporate Credit Quality
196
Basis points
250
Six-Month Commercial Paper Rate
Less Six-Month Treasury Bill Rate
200
150
100
50
1960
64
76
68
80
84
88
92
Percent of rated issues
Credit Rating Changes for Short-Term Corporate Obligations
Upgrades
-8
1973
75
77
79
81
83
85
87
Source: FRBNY Market Reports and Moody's Investor Service.
89
91
funds and the total amount issued over the last few years. Mutual fund holdings of tier
2 paper fell from a high of $25 billion to about zero at year-end 1992. The aggregate
amount of second-tier paper fell from $ 100 billion to $37 billion outstanding at year-end
1992. These declines probably reflect the credit quality concerns of investors as much
as the direct impact of the new regulations.31
The spread between top-rated commercial paper and treasury bills was never particularly large during the current recession. However, the spreads on A2/P2 paper relative
to Al/Pl paper were unusually high during 1990 and 1991, with dramatic year-end
spikes, reflecting the desire of investors that such risky holding not show up on their
public accounting statements (see Chart 14). The spread itself may understate the true
extent of credit quality concerns because second-tier issuers are often "rationed" out of
the market before they drive up rates. As a result of this investor flight to quality, lower
grade commercial paper issuers exited the market in 1990 and 1991 and likely found financing available at commercial banks. Their increased demand for bank loans, however, may have caused other potential borrowers, further down the credit quality
spectrum, to be turned away by the banks.
The Public Bond Market
In the late 1980s, bond issuance soared while net equity issuance was negative for most
of the decade because of the huge amounts of equity retired through mergers and acquisitions, leveraged buyouts and stock repurchases. During the credit crunch, the market
for publicly placed investment-grade bonds demonstrated remarkable strength despite
the slowdown in economic activity and the surge in defaults and credit rating down-
31
This topic was first examined comprehensively by Mitch Post and Lcland Crabbc, 'The Effect of SEC
Amendments to Rule 2A-7 on the Commercial Paper Market," Finance and Economics Discussion Series
Working Paper #199, Federal Reserve Board, May, 1992.
Chart 13: Second-Tier Commercial Paper Held by Money Market Mutual Funds and the
Size of the Second-Tier Market
Billions of dollars (end of period)
140
• Total Outstandings
• Money Fund Holdings "
120
100
80
60
40
20
89HI
89HII
90HI
90HII
91 HI
91HII
92HI
92HII
93HI
197
Causes and Consequences
grades. In contrast, after robust growth throughout the previous decade, the market for
below-investment grade public bonds virtually disappeared in 1990 (Chart 15).32
The flight to quality that occurred in 1990 is not surprising given the relaxation of
credit standards that had occurred in the preceding years. Leverage rose and interest
coverage fell within individual S&P rating categories throughout the decade.33 With
this apparent shift in the meaning of the ratings, investors may have required higher
yields to compensate for this apparent deterioration in credit quality. Not surprisingly,
Aaa and Baa spreads over Treasuries were fairly high in the second half of the 1980s
(Chart 16). Further, credit rating downgrades exceeded upgrades for industrial bonds
by a wide margin in 1988 and 1989, in spite of the generally favorable economic environment. The surge in the ratio of downgrades to upgrades in 1990 and 1991, however,
was not unusual for a recession.
In contrast, credit to the below-investment-grade sector was sharply curtailed in
1990 and 1991 when junk bond issuance virtually halted. The credit ratios of firms issuing junk bonds had deteriorated sharply over the course of the 1980s.34 Despite this
apparent credit quality deterioration, junk bond spreads over Treasuries had remained
fairly stable through mid-1989 (Chart 17). But the junk collapsed in late 1989, following the adoption of the thrift-bailout legislation (FIRREA) in August 1989 which pro32
In a pattern similar to the junk market growth, public markets also provided a substantial amount of equity
capital to small- and medium-sized firms during the mid-1980s through initial public offerings and venture
capital funds, but dried up in 1989 and 1990.
33
Standard and Poor's CreditStats shows that pretax interest coverage deteriorated from 1987-89 to 1990-92
in all ratings categories except AAA while the debt/capitalization ratio deteriorated for all ratings except
B.
34
See Barrie Wignore, 'The Decline in Credit Quality of New-Issue Junk Bonds," Financial Analysts Journal, Sept.-Oct., 1990.
Chart 14: Quality Spreads in the Commercial Paper Markets
198
Basis points
200
150 -
100 -
A2P2 less A1P1 Commercial Paper Rates
1987
92
93
Source: Moody's Investment Service and Federal Reserve Bank of New York, Market
Reports.
hibited thrifts from investing further in junk bonds, the unsuccessful leveraged buy-out
of United Airlines in October 1989, and the failure of Drexel in February 1990. Default
rates on corporate bonds, mostly original-issue junk bonds, began to reach unprecedented heights (see Chart 17). New issuance of publicly traded junk bonds had virtually
ceased in 1990 and spreads on junk bonds soared.
In addition to directly reducing the availability of bond finance for new M&A/LBO
activity, the collapse of the junk bond may have squeezed bank credit for traditional C&I
loan customers as banks' "bridge loans" that could no longer be refinanced in the bond
market became an extra burden. Recently, the market for publicly issued junk bonds has
rebounded with low spreads over Treasuries and record issuance in 1992. For privately
placed junk bonds, however, there is still no evidence of a recovery.
Securitization of Receivables
Over the past four years, securitization continued the rapid growth that began in 1970 and
has helped relieve the capital constraints on credit growth at financial intermediaries. Despite the continued contraction of the thrift industry, residential mortgage credit remained
plentiful over the past few years largely because of the mortgage-backed securities market. In addition, the growing securitization of nonmortgage credit limited the potential
restraint on credit created by the capital shortages faced by banks, thrifts and finance
companies. Although the market is still developing, securitization of consumer and business loans has already enabled many intermediaries to raise their capital ratios and to reduce their funding requirements without necessarily cutting back on loan origination.35
35
Besides securitization, banks can sell their loans outright to banks and other investors, thereby maintaining
origination while limiting balance sheet growth. For a review of securitization and loan sales during the
credit crunch period, see Richard Cantor and Rebecca Demsetz, "Securitization, Loan Sales, and the
Credit Slowdown," Quarterly Review, Federal Reserve Bank of New York, Summer, 1993.
Chart 15: Nonconvertible Public Domestic Debt Issuance
Billions of dollars
350
300 -
Below Investment Grade
1978 79
80
81
82
83
84
85
86
87
88
89
90
91
92
Source: Morgan Stanley.
199
Causes and Consequences
At year-end 1992, home mortgage-backed securities (at $1.4 trillion) exceeded the
amount of home mortgages held on the books of depository institutions and mortgage
companies (at $1.3 trillion).36 Over the prior four years, mortgage-backed securities
grew about 70 percent ($584 billion) while home mortgages at depository institutions
and mortgage companies grew about 10 percent ($120 billion) (see Chart 18).
The rate of growth of securitization of consumer credit over the past few years has
been even more rapid than the growth of mortgage-backed securities. In total, securi-
36
We have included the home mortgage holding of government-sponsored enterprises in the stock of mortgage-backed securities because these mortgages are in the process of being packaged for sale.
Chart 16: Long-Term Corporate Credit Quality Indicators
200
Basis points
350
Investment Grade Corporate Bond Yield
_
Less Ten Year Treasury Rate
300
^ Baa Corporate rate less
10 year Treasury rate _
250
200
150
100
50
1960
72
64
76
80
84
88
92
Numbers
Credit Rating Changes for U.S. Corporate Bonds
1987
88
89
90
91
Source: Moody's Investor Service, FRBNY Market Reports.
93
tized consumer credit rose $95 billion, from $30 billion to $125 billion between yearend 1988 and 1992 (see Chart 19). Banks and thrifts have been the leading issuers of
securitized revolving credit, and finance companies have led the securitization of consumer auto loans.
Although somewhat harder to package than consumer credits, business loans are
now being securitized as well. Finance companies have begun to securitize a wide variety of business credits, including leases. These programs have increased from $1 billion to $12 billion over the past four years (see Chart 19). Asset-backed commercial
paper, in existence since 1983, is another way in which commercial credits are securitized. As shown in Chart 20, the amount of commercial paper issued by asset-backed
Chart 17: Developments in the Below-lnvestment Grade Corporate Bond Market
Percent
14
Below-lnvestment Grade Bond Yield
Less Ten Year Treasury Rate
12
Recession
10
8
6
Insider Trading Scandal
Stock
LTV Bankruptcy
Market RJR Issuance
Panic
4
2
U
1986
87
88
89
91
90
92
93
Percent
2.0
Corporate Bond Defaults
4.0
IV
Percent of Total Debt I
Outstanding
1
(left scale)
I
1.5 -1
•
-3.0
A hi
1.0 Percent of Total Number
of Issuers
(right scale)
0.5 -
|
\
1
xf
\ j \
11\^
Tr
I f
^***
\
1
1
- 2.0
- 1.0
— *
0.0
1970
72
74
76
78
80
82
84
86
88
90
0.0
92
Source: Morgan Stanley, Moody's Investor Service.
Note: Data for 1993 reflects the number of defaults through September and the dollar value
of defaults through August.
201
Causes and Consequences
programs has grown from roughly $7 billion to approximately $58 billion between yearend 1988 and year-end 1992. The commercial paper issued by these special purpose
corporations is collateralized by a diversified pool of credits, often business trade receivables. When this asset-backed structure is combined with some other form of credit support, such as a third-party guaranty, the commercial paper often receives a top rating
from the credit rating agencies. Banks or investment banks that arrange these programs
for their clients are, in effect, using the capital markets to provide inexpensive funding
to businesses that cannot access the markets directly, as cheaply, or at all.
The extent to which securitization has relieved the pressures on financial intermediaries can be measured in part by the net increase of $635 billion in asset-backed securities over the last four years. However, the intermediaries that do securitize their assets
Chart 18: Home Mortgage Funding
202
Billions of dollars
1,600
Securitized Mortgages
1,400
12 Government Sponsored Agencies
.
• Mortgage Pools
1,200
1,000
800
600
400
200
0
1985
86
87
89
90
91
92
Billions of dollars
1,400
Mortgages Held by Intermediaries
1,200
1,000
• Finance and Mortgage Companies
0 Commercial Banks
. • Thrifts
800
600
400
200
1985
87
89
91
Source: Flow of Funds, Board of Governors of the Federal Reserve System.
93
may be those most under financial stress. Hence, securitization may have had a larger
impact on credit availability than suggested by the aggregate data. The asset-backed securities market enabled many financial intermediaries to continue their function as "loan
originator" while passing on their role as "asset holder" to individual investors and better capitalized institutions.
IV. Summary
This paper provides a quantitative analysis and qualitative discussion of the recent behavior of the major sources of nonbank business credit—loans by finance companies,
commercial paper issuance, and bond issuance. The unusual weakness in nonbank
Chart 19: Asset-Backed Securities Outstanding
Billions of dollars
uu
Consumer Credit
-•
—II i j
• Securitized by Finance Companies
80 " • Securitized by Banks and Thrifts
60 -
40 -
20
n
1988
89
90
91
92
91
92
Billions of dollars
14
Finance Company Business Credit
1988
89
Source: Federal Reserve Board.
203
Causes and Consequences
credit over the 1989-1992 can be traced to (1) financial distress at many finance companies and life insurers, (2) a rash of downgrades and defaults in the junk bond and commercial paper markets, and (3) changes in attitudes by investors and regulators. These
factors and similar developments occurring at the same time in the banking system likely
reduced credit availability for below-investment-grade and unrated commercial borrowers.
Nonbank business credit began to slow sharply beginning in late 1989, in parallel
with the decline in bank credit. This pattern contrasts sharply with that observed in past
credit crunches in which the growth of nonbank credit accelerated (at least briefly) to
offset a decline in bank credit. Our analysis of the time series data finds (1) little evi-
Chart 20: Asset-Backed Commercial Paper
204
Number of programs
140
120 100 80 60 -
Billions of dollars outstanding
1986
87
88
89
90
91
Source: Calculations provided by Mitch Post of the Federal Reserve Board.
92
dence of a shift from bank to nonbank sources of funds during the current slowdown and
(2) the declines in bank and nonbank credit growth far exceed the amount predicted by
their historical relationships with the business cycle-related determinants of credit demand.
On previous occasions, the precipitating factor behind a credit crunch was disintermediation; whereas, in the recent episode, the primary causes of the declines in bank and
nonbank credit growth are similar. In late 1989, the junk bond market collapsed following a rise in bond defaults and a failed attempt to buy United Airlines. At the same time,
large losses announced by the Bank of New England, among others, made the public
aware that the commercial real estate bubble had already burst. Like numerous banks,
many finance companies and insurance companies suffered poor performance in loans
to highly leveraged companies and in commercial real estate mortgages combined with
insufficient capital to cushion against the related losses. Moreover, again like banks, finance companies and life insurance companies were pressured by the actions of the
credit rating agencies, private investors, and regulators to improve asset quality,
strengthen capital bases, and limit balance sheet growth. At the same time, a rash of
commercial paper downgrades and defaults raised concern about the safety of money
market mutual fund investments, reducing the demand for lower quality commercial paper and causing the SEC to limit the money market fund holdings of lower quality paper.
Beginning in late 1989, virtually all investors shunned commercial real estate and
junk bonds. Junk bond issuance halted and the yield spread between above- and belowinvestment-grade bonds widened dramatically. The private market for below-investment-grade bonds was particularly hard hit as its major investor, life insurance companies ceased purchasing. In 1990 and 1991, both banks and finance companies raised
their credit standards and widened their lending spreads. Commercial paper issuance
by firms that lacked top credit ratings fell dramatically. As a result, below-investmentgrade and unrated firms experienced a decline in credit availability across a wide spectrum of funding sources.
Nonfinancial firms with strong credit ratings, however, had uninterrupted access to
credit in the money and capital markets—as investment-grade quality yield spreads,
both at short and long maturities, remained relatively flat throughout the recession.
Moreover, the market for high quality fixed-income securities remained strong, and
many financial intermediaries were able to issue asset-backed securities that relieved, in
part, their balance sheet constraints and allowed them to continue lending.
Appendix. A Review of Sources of Business Credit
Among the various types of business credit, commercial bank lending is unique in the
breadth of its customer base (across firm size and credit quality) and the flexibility of its
terms. For large corporations, however, borrowing in the direct capital markets, through
bond or commercial paper ("CP") issuance, is generally less expensive compared to
funding through commercial and industrial ("C&I") bank loans. Small and mediumsized firms, without direct access to the capital markets, generally borrow from commercial banks, but they do have access to other important sources of credit. In particular, they often borrow from finance companies, whose business loans are usually more
specialized and require more collateral than traditional C&I loans. Another important
source of funding for small- and medium-sized firms is interfirm trade debt which enables the business sector as a whole to reduce its external borrowing needs. Some less
quantitatively significant sources of funds—such as venture capital funds and business
lending by thrifts—may also be of particular importance for the small business sector.
205
Causes and Consequences
Table Al provides estimates of selected liabilities of the nonfinancial business sector.
Commercial and Industrial Bank Loans
Commercial banks provide business credit to a wide variety of borrowers, ranging from
small start-up companies to the largest corporations. U.S. money center banks and U.S.
branches and agencies of foreign banks specialize in the provision of credit to large corporations and partnerships. Borrowers with strong credit ratings, however, generally
meet most of their funding requirements in the direct capital markets and tend to borrow
from banks only on a short-term basis for temporary liquidity. Large corporations with
lower credit ratings, however, rely more heavily on large banks to meet working capital
needs and to provide medium-term financing for capital expansion or financial restructurings. Some loans are too large to be financed by one bank alone and are funded jointly by many banks through loan participations which are sometimes traded in the
secondary markets.
Regional banks may participate in the large credits described above, but they also
tend to specialize in middle-market lending, funding working capital needs or capital
expansion programs of medium-sized companies. In addition to regional banks, community banks also lend to small businesses through traditional C&I loans or with Small
Business Administration loan guarantees. Bank credit is normally extended only to
small and medium-sized firms that (1) possess strong balance sheets and collateral, and
(2) can present evidence of a strong credit history and banking relationships. Lesser
quality credits are more likely to be able to obtain funds from nonbank lenders.
The terms to maturity of individual C&I loans vary widely, ranging from overnight
Table A1: Selected Liabilities of the Nonfinancial Business Sector
Billions of Dollars
206
Bonds
a
Privately placed bonds b
Bank C&I loans
a
Finance company loans
Commercial paper
Thrift C&I loans
a
8
3
Venture capital funds0
Trade debt
3
Securitized business credit
d
Year-End 1992
Year-End 1988
Percent Change
$1,231
$969
27
250
n.a.
n.a.
661
670
-1
297
245
21
108
86
26
8
34
-76
35
n.a.
n.a.
811
686
18
64
8
689
Sources:
- Board of Governors, Flow of Funds.
b
Estimate for 1991 found in Mark Carey, Stephen Prowse, John Rea, and Gregory Udell,
"Recent Developments in the Market for Privately Placed Debt." Federal Reserve Bulletin,
February 1993.
c
- Steven Bavaria, "Will Venture Capital Bloom in the 1990s?" Investment Dealer's Digest,
May 25, 1992.
d
- Federal Reserve Bank of New York estimates of the sum of securitized finance company
loans and leases and asset-backed commercial paper supporting business credit.
a
(e.g., for cash management), to under a year (e.g., for working capital), to multi-year
(e.g., for equipment purchases or LBO-related financings). Moreover, many C&I loans
have no explicit maturities, rather they are extended under lines of credit that can be
drawn down on demand by the borrower, and perhaps can be called on demand by the
lender as well.
The majority of bank C&I loans are at least partially secured by the borrowers' assets. The details of the collateralization, however, are often left imprecise and the bank
credit review process traditionally decides on the acceptability of a particular loan application as if the loan were to be made on an unsecured basis. This informal method
of collateralization and greater stress on balance sheet/cash flow considerations contrasts strongly with secured loans made by finance companies. Finance companies typically look much more closely at the quality of the collateral as a basis for credit
approval.
Commercial Paper
Large firms with strong credit ratings often have commercial paper ("CP") programs
which provide funds for their daily cash management and working capital needs. To be
exempt from registration requirements, the SEC requires that CP mature in less than 270
days and be issued for working capital purposes such as inventory or accounts receivable financing. In practice, almost all CP issuers have investment-grade credit ratings
and have back-up bank credit facilities for their commercial paper programs. The largest purchasers of CP are money market mutual funds. In addition, commercial paper is
held by pension funds, insurance companies, households, and corporate accounts managed by bank trust departments.
Commercial paper issuance is highly concentrated among a small number of firms,
most of which have "Tier 1" credit ratings (i.e., rated P-1 by Moody's or A-1 by S&P or
D-l by Duff and Phelps or F-l by Fitch) Smaller firms or large firms with weak credit
ratings generally do not have access to the commercial paper market. Recently, however, special purpose funding vehicles have been set up by bank holding companies and
securities firms to issue highly rated CP often backed by pools of trade receivables of
firms that do not, themselves, have top credit ratings. The total amount of receivables
outstanding that are currently securitized in this manner is about $58 billion.37
Long-Term Capital Markets
For large corporations, long-term funding for business expansion or general corporate
purposes is generally obtained in the public bond and equity markets. For small and medium sized businesses, venture capital funds and initial public offerings of equity are important sources of risk capital (see Chart Al). Prior to the 1980s, access to the public
bond market was generally available only to firms with investment-grade credit ratings.
The subsequent development of the junk bond market, however, extended market access
to the below-investment-grade sector as well. This sector now represents almost 20 percent of the public debt market. Net issuance of publicly traded bonds was extremely
rapid during the 1980s, but much of the funds raised with bonds was used to retire corporate equity.
The private placement bond market competes more directly with banks for market
37
An excellent introduction to this topic has been written by Barbara Kavanaugh, Thomas Boemio, and Gerald Edwards, Jr., "Asset-Backed Commercial Paper Programs," Federal Reserve Bulletin, Vol. 78, Num.
2, February, 1992, pp 107-116.
207
Causes and Consequences
share in lending to middle market corporations. Medium sized-firms with moderate
funding needs and varying credit quality often either cannot or choose not to borrow in
the public markets and instead obtain long-term and medium-term funding in the private
placement market. Bonds and stocks that are sold privately are not subject to the Securities and Exchange Commission's registration requirements and disclosure rules, which
require that borrowers report actual and potential obligations affecting their ability to repay debt. By issuing privately, companies can save the direct and indirect costs of these
requirements. Privately placed securities have disadvantages, however, in that the SEC
restricts the extent to which they can be resold to other investors. Because they are
Chart A1: Trends in Venture Capital and IPOs
208
Billions of dollars
5
Capital Commitments to Venture Capital Funds
1979 80
81
82
83
84
85
86
87
88
89
90
91
92
84
85
86
87
88
89
90
91
92
93
Billions of dollars
40
Initial Public Offerings
30
20
10
1980 81
82
83
Source: National Venture Capital Association.
therefore less liquid, private placements must typically offer investors a compensating
premium in yield relative to public securities for the borrowers of the same risk class.
The primary investors in the private placement market are life insurance companies
and a few pension funds. The market is like an intermediated credit market in that the
lenders need to perform their own credit evaluation of the borrower and tend to monitor
closely his financial condition over time. At year-end 1991, life insurers held about
$212 billion in privately placed corporate bonds. At the time, the total stock of privately
placed, nonfinancial corporate bonds outstanding was about $250 billion.38
Life Insurance Companies
Life insurance companies are major suppliers of business credit through their purchases
of foreign and domestic corporate bonds and commercial real estate mortgages. At yearend 1992, life insurers held $650 billion in foreign and domestic corporate bonds and
$250 in commercial mortgages. During 1990 through 1992, insurers' bond holding grew
roughly 8 percent per year while their holdings of commercial mortgages remained virtually unchanged. These patterns of growth were quite similar to the growth rates observed over the same time period for corporate bonds and commercial mortgages in the
aggregate. Life insurers, however, serve a key role in financing business credit because
they, like banks, specialize in providing credit to borrowers that cannot easily obtain
funds from the direct credit markets. In particular, of the $250 billion privately placed
bonds that have been issued by U.S. nonfinancial corporations, about two-thirds have
been purchases by life insurers. Moreover, of this roughly $165 billion in private placements, about 17 percent are issued by below-investment-grade borrowers.
Finance Companies
Finance companies are major providers of business credit to a wide variety of firms.39
The lenders in this industry are extremely diverse and include (1) so-called "captives",
subsidiaries of manufacturing companies that mainly provide sales finance, leases, and
dealer loans for their parents' products, (2) large diversified lenders that often are subsidiaries of industrial firms, financial firms, or utilities and provide a wide array of services, ranging from factoring, to small and large ticket leasing, to mergers and
acquisition financing, (3) independent companies often specializing in either small business finance and/or leasing.
The traditional customer base of finance companies consists of small and middle
market companies, with perhaps lower than average credit quality and without access to
bank credit. The captive finance companies, in part to meet the sales objectives of their
parents however, often compete with banks for the business of lending to borrowers of
higher credit quality . Moreover, independent finance companies often compete successfully with banks for financing large investment grade borrowers when the funding
arrangement requires highly specialized "niche" expertise, such as knowledge of the aircraft or other "large-ticket" leasing businesses.
38
These estimates, as well as a wide ranging discussion of the private placcmcnl market, arc contained in
Mark Carey, Stephen Prowse, John Rea, and Gregory Udell, "Recent Developments in the Market for Privately Placed Debt," Federal Reserve Bulletin, February, 1993.
39
The finance company industry is surveyed in a paper by Eli Remolona and Kurt Wulfekuhler, "Finance
Companies, Bank Competition, and Niche Markets," Quarterly Review, Federal Reserve Bank of New
York, Summer, 1992, pp. 25-38.
209
Causes and Consequences
Trade Debt
210
The interfirm transactions that support the production and merchandising of goods are
facilitated by a substantial volume of trade debt and trade credit. At year-end 1991, outstanding trade debt was about $780 billion, roughly equal to the total volume of outstanding C&I loans. In the aggregate data, trade debt tracks the business cycle and
inventories, in particular, very closely. As a result, fluctuations in trade debt may reveal
only scant independent information about economic activity or credit conditions.40 A
closer examination of the data, however, can reveal other patterns. Firms in the wholesale trade sector and small manufacturing firms tend to rely more on trade debt than other firms as a means of inventory finance.41 Businesses that lack current funds but need
to acquire raw materials, intermediate goods, or inventory for final sale can, within
bounds, increase their credit by "stretching-out" their payables.
40
Only a few macrocconomists have studied trade credit in great detail, either from theoretical or empirical
perspectives. A recent paper by Valerie Ramey examines the correlations between fluctuations in trade
credit and the monetary aggregates and cites the related literature. See her paper, "The Source of Fluctuations in Money: Evidence from Trade Credit," NBER working paper #3756, June, 1991.
41
Sec the Quarterly Financial Report for Manufacturing Corporations, Department of Commerce, various
issues.
Survey Evidence on Credit Tightening and the
Factors Behind the Recent Credit Crunch
by Kausar Hamdani, Anthony P. Rodrigues, and Maria Varvatsoulis]
Media reports that some firms were having trouble obtaining loans, coupled with a sharp
and persistent decline in bank loans to business (Chart 1), have led some commentators
to suggest that a credit crunch occurred during 1989-92. However, the drop in bank
lending could have been caused by a decline in either loan demand or loan supply, arising from the recession and subsequent slow economic growth. Surveys of borrowers
and lenders can indicate the relative importance of demand and supply factors for the
decline in business loans. In this paper, we analyze both survey evidence and macroeconomic data to investigate the effects of supply and demand factors on the observed credit tightness.
Other researchers have also examined survey evidence to explain credit developments in the recent period. In contrast to much of this literature, which focusses on surveys of either borrowers or lenders, we examine both borrower and lender surveys for
evidence of credit tightness.2 We also analyze large and small business borrowers, reviewing the extent and the form of the credit tightening that both classes of borrowers
experienced. Where possible, we compare the current period with historical episodes.
The paper is structured as follows: The first and second parts of the paper are narrative analyses of credit conditions in the mid-1980s and in 1989-92. The later analysis
suggests that, while demand factors contributed to the 1989-92 slowdown, supply factors played an important role, particularly in 1990, and that the tightness faced by small
businesses has been longer-lived than the credit restraint on large firms. In the third and
fourth parts of the paper, we motivate and use regression models to purge the survey
variables of their cyclical component, providing noncyclical measures of credit supply.
1
The authors wish to lhank Akbar Akhtar, Richard Davis, Valerie LaPorte, Patricia Mosser, Charles Stciridel, and John Wenninger for helpful suggestions on earlier drafts of this paper.
2
For example, S. Schrcft and R. Owens, in "Survey Evidence of Tighter Credit Conditions: What Does It
Mean?" Federal Reserve Bank of Richmond, Economic Review, March/April 1991, focus on lender surveys while W. Dunkelberg, in "Small Business Credit Crunch," The NFIB Foundation, December 1992,
examines surveys of small business borrowers.
211
Causes and Consequences
In the fifth part of the paper, we introduce survey measures of credit supply into loan
growth models. Both the survey and loan growth regression models suggest that tight
credit supply contributed to the recent weakness in loan growth. Finally, a description
and history of the major systematic surveys is provided in the Appendix.
I. Survey Evidence for the Mid-1980s: The Benchmark Period
Nonfinancial business credit boomed in the mid-1980s (Chart 2). Bank lending to businesses also grew quickly in this period, although not at the pace of other components of
credit. The demand for debt was driven by merger and acquisition activity and corporate
restructuring. At the same time, financial innovations and the deregulation of financial
institutions supported the supply of credit.
Three major surveys provide insight into bank loan behavior. Small business borrowers are surveyed by the National Federation of Independent Business (NFIB). 3 Bank
lenders are represented by the Federal Reserve's Senior Loan Officer (SLO) survey.4 The
perceptions of large and middle market business borrowers are recorded in the Goldman
Sachs & Co. survey of Fortune 1000 chief financial officers (CFOs).5
3
The National Federation of Independent Business (NFIB) has surveyed small businesses about credit availability and business activity since late 1973. Since small businesses rely heavily on bank credit, their perceptions of credit conditions most likely reflect bank loan availability.
4
The Federal Reserve System has conducted the Senior Loan Officer Opinion (SLO) survey since 1964.
The information elicited is primarily qualitative and indicates the loan policy stance of each participating
bank. If respondents have changed their loan policies, they are asked to report the direction and degree of
the change. Questions on credit availability and loan demand were asked only occasionally during the
mid-1980s.
5
Goldman Sachs & Co. has surveyed the chief financial officers (CFOs) from the Fortune 1000 industrial
Chart 1: Commercial and Industrial Loans at U.S. Commercial Banks
212
Billions of dollars (1987)
700
600
500
400
300
200
1967 69
71
73
75
11
79
81
83
Source: Flow of Funds, Bureau of Economic Analysis.
85
87
89
91
93
Large and Middle Market Business Borrowers
The SLO surveys of banks during the mid-1980s suggest that loan growth was mainly
demand driven and that insufficient supply of bank credit was not a major problem.
Weak demand was mentioned more often than tighter standards as a source of slow loan
growth in the surveys from 1984 to 1988 (Table 1). When weakness in loan demand existed, it was largely attributed to declines in firms' financing needs, although the increased use of nonbanks and capital markets was also occasionally cited as a factor
(Table 2). Furthermore, in 1986 a majority of banks reported undertaking several marketing changes specifically to increase loan issuance, suggesting that supply was supporting loan growth.
The CFO surveys show that the fraction of large businesses planning to increase
bank loans during 1985-88 averaged slightly below the levels of 1976-83 (Chart 3). Using reported CFO expectations of growth in inventories, capital spending, and cash flow,
we can construct a rough proxy for net demand for external funding since 1985 (presented in Chart 3). The relatively close association between the proxy for demand for external funds and the fraction of Fortune firms intending to increase their borrowing
suggests that any weakness in loan growth over 1985-88 was mainly demand driven.
Decomposing the respondents by size and industry, we find that the weakness in demand
during 1985-86 was pervasive with demand strengthening later in the period, particularly in 1987 (Chart 4).
Footnote 5 continued
companies semiannually since 1976. A question on expected borrowing from banks, for the upcoming six
months, relative to the previous six months, has been asked consistently since the inception of the survey.
Chart 2: Nonfinancial Business Credit
Relative to Business GDP
Percent
80
1967 69
71
73
75
77
79
81
83
85
87
89
91
93
Source: Flow of Funds, Bureau of Economic Analysis.
213
Causes and Consequences
Small Business Borrowers
The NFIB surveys indicate that small businesses shared in the general economic prosperity of the mid-1980s. From 1986-88, a growing fraction of firms reported higher sales
and increasing inventories (Table 3). This economic buoyancy implied healthy small
business loan demand.
Supplementary survey data for this period are limited. One 1988 survey, the NAM
Small Manufacturers Operating Survey, reports a large fraction of firms experiencing
better sales and suggests a relatively strong loan demand over 1987 and early 1988 (Table 4). 6
6
The National Association of Manufacturers surveys its members early in the spring of each year. Since the
credit availability question asks respondents to compare current conditions with those in the prior year, the
data likely reflect changes in conditions during the preceding calendar year.
Table 1: Reports of Weaker C&l Loan Growth
Senior Loan Officer Survey, Percent of All Banks
214
Sources of Weaker C&l Loan Growth
Survey Date
Weaker Business Loan
Demand
Percent
Tighter Credit Standards or
Decreased Willingness to
Lend
Banks
Reporting
Weaker C&l
Loan Growth
Net Percent
Percent
Net Percent
Percent
-
27a
5
7
-
43
25
1984
Jun.
Sep.
Nov.
37
22
1985
Feb.
27
-
5
(3)b
30
1986
Aug.c
7
(16)
4
(16)
11
1987
Sep.
42
34
5
-
53
1988
Nov.
28
-
0
-
28
1989d
33
-
26
-
51
1990
May
27
•
24
-
44
Source: Federal Reserve System.
a
- Percent of banks reporting tighter standards, but not attributing weaker loan growth to them.
b
- Net percent of banks reporting a decreased willingness to lend, but not attributing weaker loan
growth to it.
c
- For middle market borrowers only.
d
- Information regarding 1989 was provided in the May 1990 survey.
Credit supply conditions were neutral for small businesses during the mid-1980s.
Over 1985-88, the net percent of NFIB regular borrowers who reported that credit was
more difficult to obtain increased slowly, remaining below the series' average value
(Chart 5). Financing costs were not a sustained problem because interest rates declined
on average following the 1982 recession even though a significant fraction of borrowers
reported a higher loan rate in 1984 and 1987.
The NFIB survey indicates, however, that credit supply conditions worsened in 1988
(Chart 5). Concurrently, the percent of NFIB firms reporting a higher borrowing rate increased sharply, suggesting that rising loan rates were the main reason for the growing
reported incidence of credit stringency.
Table 2: Sources of Large and Middle Market Firm Weaker Loan Demand
Senior Loan Officer Survey, Net Percent of All Banks
Firm Size/
Survey Date
Large
Medium
Substitution to Nonbank Financing
Medium
Large
1984
Apr.
Sep.
Nov.
(42)
18
17
(13)
17
14
1985
Feb.
15
8
1987
Sep.
19
17
1988
Nov.
17
0
1989
33
21
1990
May
22
27
11
1991
Aug.
Oct.
16
25
10
31
CM CO
Reduced Funding Needs
1992
Jan.
May
Aug.
Nov.
24
2
10
2
31
(9)
3
(2)
10
15
5
0
0
0
0
3
1993
Feb.
May
Aug.
Nov.
(11)
0
(7)
(3)
(16)
(5)
(16)
(10)
(5)
18
4
8
0
4
0
3
5
(2)
3
Source: Senior Loan Officer Survey, Federal Reserve System.
Note: As of 1990, banks have responded separately for large firms (those with sales greater
than $250 million) and middle market firms (those with sales between $50 million and
$250 million).
215
Chart 3: Indicators of Bank Loan Demand
Goldman Sachs Survey of Fortune 1000 CFOs
Net Percent of Firms
Percent
5
50
Firms planning to increase bank
borrowings in the ensuing six months
^
(left scale)
25
-5
Demand proxy
(right scale)
-25
1976
90
78
92
93
-10
Source: Goldman, Sachs & Co.
Note: The correlation between the two series is 0.70 in 1985H1 to 1990H1, and 0.21 in
1990H2 to 1993H1. The proxy for external funds is capital spending plus inventory growth
minus internal cash flow growth, all projected over the next six months.
Chart 4: Large and Middle Market Firm Bank Loan Demand
Goldman Sachs Survey of Fortune 1000 CFOs
Net Percent of Firms
60
40
20
-20
-40
1976
78
82
84
86
88
90
92
93
Source: Goldman, Sachs & Co.
Note: Firms planning to increase bank borrowing in the ensuing six months.
216
II. Survey Evidence for 1989-92
The 1989-92 credit slowdown can be divided into two subperiods: 1989-91 and 1992.
The first represents a period of credit weakness and decline; the second coincides roughly with the credit trough. The first period began with softening loan demand in 1989
while lenders, responding to deterioration in the quality of loan portfolios, became more
cautious. During 1990-91, both loan demand and credit availability dropped relative to
1989 and the mid-1980s. By mid-1992, loan demand began to recover and credit conditions stabilized with the absence of further tightening in credit availability.
Large and Middle Market Business Borrowers
Loan Demand
SLO surveys of banks reported shrinking demand from middle market and large business borrowers between 1990 and 1992, with more pervasive demand weakness for
larger customers (Chart 6). This weakness was attributed primarily to reduced funding
needs, though large firms' tendency to substitute to nonbank sources of funds7 was also
noted (see Table 2).
By the end of 1992, the SLO surveys reported that loan demand was recovering at
middle market and large corporate borrowers. The demand at mid-sized firms seemed
driven primarily by the need to finance inventories and plant and equipment, while demand at large borrowers arose from a shift toward bank financing from nonbanks and
the capital market.
Confirming the evidence from the SLO surveys, anticipated bank borrowing by large
7
These alternative funding sources include commercial paper, junk bonds, and investment grade bonds.
Table 3: Indicators of Small Business Economic Activity
NFIB Survey, Net Percent of Firms Noting an Increase
Year
Net Earnings
Sales
Inventory
Capital Outlays8
1986
(10)
7
1
55
1987
(7)
10
3
55
1988
(7)
11
5
55
1989
(13)
9
1
56
1990
(18)
4
(3)
56
1991
(25)
(4)
(10)
53
1992
(17)
3
(6)
52
1993
(21 ) b
1b
(1) b
60 c
Source: National Federation of Independent Business.
a
- Percent of firms making capital expenditures in the previous six months, as of the first month
of the quarter.
b
- Based on data available through the second quarter only.
c
- Based on data available through the third quarter only.
217
Causes and Consequences
and mid-sized Fortune companies weakened from 1989 to 1992, albeit erratically. The
CFO measure of net demand for external financing also dropped strongly (see Chart 3).
This retrenchment in borrowing plans was often stronger at the larger firms (see Chart 4).
Unfortunately, the two CFO survey indicators are not entirely consistent on the pattern of business loan demand during the period. Both indicators of demand showed declines throughout 1990. However, movements in the first indicator, the net percent of
firms planning to increase their bank borrowing, coincided with the general expectations
about the economy's behavior. The indicator rose sharply in the first half of 1991, paralleling the optimism surrounding the recession's apparent end. It subsequently fell in
the second half of 1991, when the economy appeared to have stagnated. A slight recovery appeared in the first half of 1992, when the economy seemed to be slowly growing
again. More recently, the series dropped in late 1992 and early 1993 as the slow-tomoderate pace of the recovery became evident. By contrast, the second indicator of demand, the proxy for net demand for external funding, shows no such recovery, running
below 1987-89 rates through 1991 and 1992. 8 One interpretation of the inconsistency
between the two CFO survey indicators is that the actual weakness in loan growth that
appeared in 1990-92 arose largely from demand, and that the optimism exhibited in the
borrowers' stated plans in the first half of 1991 was premature. The subsequent realignment with the downward trending proxy for net demand suggests demand continued to
be the impetus behind borrowing plans.
Loan Supply
The supply of credit to large and middle market borrowers appears to have tightened
during 1990-92. By the second quarter of 1990, a large fraction of the SLO banks re8
The lower levels of the proxy were driven by expectations of both slowing expenditures and growing cash
(low.
Table 4: Indicators of Small Business Economic Activity and Credit Availability
NAM Survey, Net Percent of Firms Noting an Increase
218
Actual
Date of
Surveya
Expected
Sales
Capital
Spending
Debt to
Cash Flow
Credit
Stringency
Sales
Capital
Spending
1988
55
-
(9)
20
54
-
1989
40
45
(5)
(8)
(60)
-
1990
5
28
(24)
(2)
47
14
1991
(12)
7
4
18
(8)
1
1992
-
-
11
5b
39
17
Source: National Association of Manufacturers.
Note: Net percent is the percent of firms reporting an increase minus the percent of firms reporting a decrease, relative to the prior year.
a
- Because the survey is conducted in February of each year and asks about conditions relative
to a year ago, it likely reflects conditions during the prior calendar year.
b
- Percentage of firms which reported not successfully obtaining credit.
ported additional tightening of standards; this tightening continued through early 1992,
although it declined in degree overtime (Chart 6). By 1992, the SLO surveys show that
any further tightening for large customers had virtually ceased. Mid-sized customers,
however, continued to face further tightening through mid-1992.
The main reason given by bank SLO respondents for the stricter credit standards was
a less favorable economic outlook, followed by industry-specific problems (Chart 7).
The other reasons cited by the respondents varied in prevalence and importance. Capital
pressures were almost as important as loan portfolio quality during 1989 and 1990.
These pressures declined in 1991 and were hardly mentioned or ranked subsequently.
Regulatory pressures, which had received prominent media attention, were cited almost
as frequently as industry specific problems during 1990, but were subsequently most of-
Chart 5: Credit Availability to Small Firms
NFIB Survey
Net Percent
\^s
\\\v
\\w\s
w
Borrowers Reporting Firmer Credit Standards than Previous Quarter
60
40
20
i 1
Ii
w
w
i
/
I f i|fV / l A r * ^
0
on
1H
I
1974
76
78
80
82
I
84
86
88
90
92 93
Percent
100
80
Borrowers Reporting Higher Lending Rates than Previous Quarter
-i
/S/IJAls 1 in w
•
60
40
20
n
1974
76
// 11
"I
78
80
82
84
86
88
90
92 93
Source: National Federation of Independent Business.
219
Causes and Consequences
ten ranked lower than both the adverse outlook and the industry problems.
Consistent with the SLO surveys, the Fortune CFO responses to questions about
bank credit tightening, initiated in October 1990, indicate that credit stringency steadily
worsened through 1991 (Table 5). Even after the onset of easing, 30 percent of all large
business borrowers still reported some form of credit tightening as of October 1992. 9
The perceived credit tightening reported by the CFOs varied somewhat by size of
firm. For most of the period, the Top 500 industrial firms experienced tightening a little
less frequently than the second 500 firms. However, this pattern was reversed in October
1992, when fewer mid-sized firms than large firms reported tightness.
9
This question was not asked in the April 1993 CFO survey.
Chart 6: Sources of Business Loan Growth
Senior Loan Officer Survey
220
Net Percent of Banks (Weighted)
80
Tighter Credit Standards
60
40
Large firms
20
Small firms
0
-20
1990
91
92
93
Net Percent of Banks (Weighted)
40
Weaker Loan Demand
20
Middle market firms
-20
Small firms
-40
1990
91
92
Source: Senior Loan Officer Survey, Federal Reserve System.
93
The form of tightening appears to have shifted during the period. Initially, stiffer
nonprice standards were more prevalent than higher pricing. Beginning in 1991, however, tighter price terms became more common as the credit stringency worsened. This
confirms the evidence from the SLO surveys, which agree that the form of tightening
shifted from stricter nonprice to higher price terms of credit, particularly for middle
market firms in late 1990 and early 1991 (Charts 8 and 9). 10 Unfortunately, more recent
10
The Fortune CFO data also indicate that, during 1990 and 1991, larger businesses encountered tightening
of price terms more frequently than smaller Fortune firms, who more often faced tougher underwriting
Chart 7: Reasons for Credit Tightening
Senior Loan Officer Survey
Percent of All Banks
70
• Capital pressures
• Loan portfolio quality
• Regulatory pressures
60
50
40
30
20
10
IV
II
1989
I
III IV
1990
III
1991
IV
I
II III IV
1992
I
II III IV
1993
Percent of All Banks
70
60
• Economic outlook
• Industry specific problems
K
50
40
\
i Banks tightening
30
20
-•
V
10 -
1
IV
II
1989
III IV
1990
ill
I
II III
1991
IV
I
I
III
1992
IV
I
II III IV
1993
Source: Includes all banks tightening standards to either large, middle market, or small firms.
Note: Senior Loan Officer Opinion Survey, Federal Reserve System.
221
Causes and Consequences
CFO information on the form of tightening was unavailable because Goldman Sachs
dropped the question from the survey in 1992.
Additional information about borrowers' perceptions of credit supply can be extracted from the question on bank borrowing plans. The drop in expected borrowing in the
first and second halves of 1990 is consistent with tightening supply of credit. Although
the sharp increase in plans in the first half of 1991 was probably motivated by economic
optimism, it suggests that either borrowers no longer perceived credit tightness or they
expected the tightness to be short-lived and unrestraining. Some analysts interpreted the
declines in borrowing plans apparent in the second half of 1991 as evidence of a credit
crunch,11 but the proxy for net demand suggests that a lack of a need for external funds
justified the drop in plans (Chart 3). Furthermore, if borrowers perceived a credit
crunch, it was not sufficient to preclude them from stating that they expected to increase
Footnote JO continued
standards. This behavior may reflect a variety of factors such as banks' attempts to retain their largest customers because of risk considerations, the borrowers' ready access to alternative sources of funds, and the
profitability of the overall banking relationship.
1[
Goldman Sachs & Co., "The Survey of Fortune Chief Financial Officers," Oct. 1991, pg. 3.
Table 5: Large and Middle Market Firm Credit Availability
Goldman Sachs CFO Survey, Percent of Firms Reporting Tight Credit Standards
222
1992
1991
1990
Oct.
Apr.
Oct.
Apr.
Oct.
46
48
51
39
30
First 500
industrials
43
48
49
35
32
Second 500
industrials
48
48
52
40
23
69
40
49
-
-
First 500
industrials
60
40
30
-
-
Second 500
industrials
77
40
68
-
-
42
63
60
-
-
First 500
industrials
53
69
67
-
-
Second 500
industrials
30
56
52
-
-
All respondents
Memo: form of tightening
Nonprice terms
All respondents
Price terms
All firms
Source: Goldman, Sachs & Co.
Note: The survey did not report the form of tightening after October 1991.
bank borrowing again in the second half of 1992.
Hence, the consistent picture the SLO and CFO surveys present of tight credit conditions for 1990-92 suggests that the credit tightness to large business evident in 1990
and 1991 was short-lived and no longer restraining by the end of 1992.
Small Business Borrowers
Loan Demand
The major influences on small business loan demand shifted dramatically over 1989-92
as initial expectations of robust growth were disappointed. Small business respondents
Chart 8: Form of Credit Tightening to Large Firms
Senior Loan Officer Survey
Net Percent of All Banks
80
Price Terms
Ill
• Cost of credit lines
• Spread of loan rates
IV
IV
1990
1992
Net Percent of All Banks
80
Non-Price Terms
1993
• Size of credit lines
• Loan covenants
• Collateral requirements
60
40
Firmer standards
20
0
-20
-40
IV
1990
IV
I
I
1991
IV
1992
I
III
IV
1993
Source: Senior Loan Officer Opinion Survey.
223
Causes and Consequences
to the NFIB survey anticipated that sales growth would be strong at the start of 1989. In
January, a large fraction of firms expected to make capital expenditures in the first six
months of the year, and more firms, on balance, expected sales and inventories to rise
than decline (Chart 10).
In the event, these expectations did not materialize. Actual earnings and sales growth
slowed during 1989, after a very weak first quarter. This sluggishness continued into
1990-92 as a shrinking fraction of firms reported improving sales, net earnings, and inventories (see Table 3). Moreover, both the NFIB survey (Chart 10) and the NAM surveys (see Table 4) show a drop in the fraction of firms planning to add to inventories or
Chart 9: Form of Credit Tightening to Middle Market Firms
Senior Loan Officer Survey
224
Net Percent of All Banks
80
Price Terms
60
•Cost of credit lines
• Spread of loan rates
40
Firmer standards
20
0
-20
-40
-60
-80
II
III IV I
I
III
1990
1991
Net Percent of All Banks
80
Non-Price Terms
IV
I
II III IV
1992
I
II III IV
1993
• Size of credit lines
D Loan covenants
•Collateral requirements
V
60 h
-
40 20
%1
Firmer standards
-i
i
-20 -40
IV
1990
I
IV
I
1991
Source: Senior Loan Officer Opinion Survey.
I
III
1992
IV
I
I
III IV
1993
to undertake new capital expenditures, particularly in 1990-91. The NFIB Small Business Optimism Index, which fell severely in 1990 and largely remained below values
observed over 1987-89, illustrates this deterioration in small business expectations
(Chart 11).
Slowing economic activity seems to have restrained small business loan demand after 1989. Although the SLO surveys report that loan demand was strong in 1989, some
weakness in loan demand developed in 1990 (Table 6). This weakness was most often
attributed to reduced needs for external funding. Substitution of funding sources either
to other banks or to nonbanks and the capital market was infrequent for small companies
who often simply reduced their financing from all sources.
By mid-1992, the SLO surveys indicated that small business loan demand had begun
to improve. Credit demand to finance inventories and plant and equipment expenditures
was the primary reason for strength, just as it had been the main source of weakness.
This recovery continued, with varying degrees of intensity, into early 1993. However,
the revival in loan demand reported by SLO lenders is not reflected in the NFIB surveys.
Loan Supply
Demand weakness was, however, not the only restraint on small business borrowing
during 1989-92. NFIB borrowers reported that credit conditions deteriorated markedly
during the first three quarters of 1989. The average interest rate paid on short-term loans
rose steadily and was on average about 1 percentage point higher in 1989 than in 1988
(Chart 12). This rise in rates was pervasive, with almost 90 percent of NFIB regular borrowers experiencing higher interest rates than in the preceding three months in 1989-H
(Chart 5). At the time of the survey, financing costs were mentioned by 11 percent of the
sample as the single most important problem facing businesses. Simultaneously, the net
Chart 10: Indicators of Small Firm Loan Demand
NFIB Survey
1986
Source: National Federation of Independent Business.
Note: Net percent of firms expecting an increase as of the first month of the quarter.
225
Causes and Consequences
percent of borrowers reporting that loans were currently harder to obtain increased
sharply, implying a tightening not seen since 1982.
By year-end 1989, the pace of tightening temporarily stabilized. Interest rates
dropped slightly and concerns about current and expected loan availability subsided to
1988 levels. The NAM surveys also reported that no noticeable credit tightening existed
as of early 1990.
However, late in 1990, the fraction of NFIB and NAM businesses reporting that
credit was harder to obtain grew again and has subsequently recovered only gradually
(Chart 12 and Table 4). This credit tightness was reported even as lending rates began
to decline. Moreover, in late 1990, NFIB borrowers expected future financing conditions to deteriorate.
The SLO survey respondents generally confirm small business borrowers' perceptions of credit stringency at least since 1989. Loan standards were tightened quite sharply in 1990 and early 1991, with additional tightening throughout the remainder of 1991
(Chart 13).12 An actual easing of standards for small business loans did not emerge until
1992. Given the modest degree of recent easing, it would seem that most of the tightening implemented during 1990-91 is still in place.
Small borrowers were more frequently subject to tightening in the form of stricter
loan covenants and col lateral ization requirements than in the form of higher spreads of
12
Overall, the main reasons for the credit tightening were the deteriorating economic outlook and industryspecific problems (see Chart 7).
Chart 11: Index of Small Business Optimism
NFIB Survey
226
Index (1978=100)
115
110 105
-
100
1975
77
81
83
85
87
89
91
93
Source: National Federation of Independent Business.
Note: Weighted average sum of the [percent favorable - percent unfavorable] responses to
questions in three categories: general expectations (expected business conditions, the climate for expansion, expected real sales volume, and expected credit conditions); current
status (current job openings, current inventory satisfaction, and change in net earnings from
prior quarter); and spending plans (plans to hire, to make capital outlays, and to add to
inventories); indexed to 1978.
loan rates over base rates (Chart 13). Nonprice tightening may have been more common
because commercial and industrial loans to small firms are often collateralized with real
estate to reflect their higher riskiness. However, when banks eased their credit standards
in 1992 for small firms, they relaxed both price and nonprice terms, offering, for example a lower cost and a larger size of credit line.
Other surveys of borrowers and lenders in 1990 also reported credit tightening for
small businesses (Table 7). A U.S. Chamber of Commerce survey highlighted that the
smallest businesses were having the most trouble. The media also carried interviews
with bankers who admitted to greater caution or "more prudential" lending practices.13
Anecdotal reports from other organizations were carried by the media in 1990 and
were consistent with the survey evidence. For example, the National Association of Investment Companies claimed that companies with sales between $5 million and $15
million "were returning to investment companies two and three times for loans, rather
than simply asking for early-stage financing. The National Minority Business Council
reported that members who had initially found longer term loans harder to get than in
the past were by mid-1990 having trouble securing even shorter term loans and working
capital. The media also carried many stories of individual firms facing more stringent
standards in the form of greater documentation, higher equity requirements, increased
collateral, and stricter loan covenants.14 Other reports told of outright denial of credit or
canceled credit lines and recalled loans. l5
In 1991, ad hoc surveys were less frequent. A Coopers and Lybrand survey reported
13
"Lenders Juggle Credit and Toxic Waste," ABA Banking Journal, July 1990, pp. 83-87.
14
Jeanne Saddler and Udayan Gupta, "Small Businesses Say They Don't Feel Credit Crunch." Wall Street
Journal May 25, 1990, pg. B2.
15
Joel Brenner, "U.S. Banks Time Lending to Consumers, Businesses," March 28, 1990. Also, Monroe
Table 6: Sources of Small Firm Weaker Loan Demand
Senior Loan Officer Survey, Net Percent of Banks
Reduced Funding Needs
Subsititution to Nonbank
Financing
1990
May
9
0
1991
Aug.
Oct.
7
17
(7)
2
1992
Jan.
May
Aug.
Nov.
16
(11)
(2)
3
(2)
0
0
0
1993
Feb.
May
Aug.
Nov.
(19)
(4)
(11)
(12)
0
2
2
0
Survey Date
Source: Senior Loan Officer Survey, Federal Reserve System.
227
Causes and Consequences
Chart 12: Indicators of Small Firm Credit Conditions
NFIB Survey
Percent
14
Average Interest Rate Paid
228
1986
89
87
90
91
92
93
91
92
93
91
92
93
Percent of Firms
14
Current and Expected Credit Stringency
12
10
Net percent expecting credit
to become harder to obtain
8
6
Net percent reporting credit
currently harder to obtain
4
1986
87
88
89
90
Percent of Firms
Financing as the Most Important Problem
1986
87
88
89
90
Source: National Federation of Independent Business.
that while 34 percent of companies were able to increase their bank borrowing in the
third quarter, only 24 percent of the respondents were able to do so in the fourth quarter.16 Firms reported that banks had raised their lending criteria by demanding higher eqFootnoie 15 continued
Karmin and Robert Black, "Ouch! The Squeeze is On," U.S. News and World Report, April 23, 1990, pp.
51-52; and Gene Koretz, "The Bank Credit Crunch Stymies the Small Startup," Business Week, October
29, 1990, pg. 22.
16
"Trendsetter Barometer," a publication of Coopers and Lybrand, the accounting firm, tracks the quarterly
performance and business concerns of over 200 growth companies. Both manufacturing and service com-
Chart 13: Form of Credit Tightening to Small Firms
Senior Loan Officer Survey
Net Percent of All Banks
ou
Price Terms
• Cost of credit lines
• Spread of loan rates
60
\
Firmer standards
40
-
ii
m
20
-Tl
-20
m-i
"•frtKll n p
f! 1
0
-n
-
-
.An
II
III IV
1990
IV
I
I
1991
Net Percent of All Banks
80
Non-Price Terms
60
II III IV
1992
I
II III IV
1993
• Size of credit lines
D Loan covenants
• Collateral requirements
\
Firmer standards
40
20
-20
I
III
IV I
II III
IV I
1990
1991
Source: Senior Loan Officer Opinion Survey.
II III
1992
IV I
II III I V
1993
229
Causes and Consequences
uity, collateral, and cash flow coverage. Fully 9 percent of all respondents reported that
their efforts to increase credit had been rebuffed by bankers. This behavior existed even
as bankers lowered the interest rate on loans.
Despite the media reports, contemporary analysts resisted diagnosing credit conditions as a "credit crunch." They were chiefly interested in whether the credit stringency
was sufficient to trigger an economic recession. For example, in 1991, Jerry Jasinowski,
President of the NAM, concluded that "the credit rationing process is limited to specific
sectors and regions, and is not a systemic or economy-wide phenomenon." l7
In addition, the tightening was hard to document as a "crunch," except in an anecdotal sense, because credit-squeezed businesses were reluctant to identify themselves,
according to the National Small Business United. They did not want their suppliers or
their bankers to know of their difficulty in obtaining credit.18
Moreover, some lenders maintained that there was no credit crunch if the term was
interpreted to mean that creditworthy borrowers were being denied credit. The weak
loan growth was instead attributed to decreased borrower creditworthiness and loan demand. In fact, many lenders contended that credit standards, as such, had not been tightened. Instead, implementation had relaxed during the 1980s and bankers were now
simply returning to earlier practices.19
Footnote 16 continued
panies are covered. Company size ranges from revenue/sales of $1 million lo $50 million, with a median
value of $6.4 million and 50 employees. These firms arc characterized as among the fastest growing U.S.
companies.
17
Jerry Jasinowski, "The Credit Crunch and the Export Surge," June 27, 1990, pg. 2.
18
Saddler and Gupta, op. cit.
19
Brian Dittcnhafer, President of FHLBNY, "Statement before the Invitational Hearing," New York Senate
Table 7: Results of Other Recent Surveys on Credit Availability: 1990
Percent of Sample
230
Percent Reporting Tighter
Credit Conditions
Connacticut Business and Industry Association May Survey
1. Percent reporting credit harder to obtain relative to one year ago
76.2
U.S. Chamber of Commerce July Survey
2. Percent reporting credit harder to obtain over the last six months
21.5
Percent of businesses with sales less than $1 million reporting credit
harder to obtain over the last six months
27.4
Percent of businesses with sales over $10 million reporting credit harder
to obtain over the last six months
12.3
ABA Banking Journal November Survey of Banks
3. Percent reporting contracting credit in their primary market
22.0
4. Percent reporting contracting credit for marginal borrowers
85.0
5. Percent reporting tightening lending standards
49.0
6. Percent reporting weak loan demand
69.0
Overview of Survey Evidence on Loan Demand and Supply for 1989-92
Surveys of lenders and borrowers agree that both loan supply and demand weakened
during 1989-1991. These weaknesses arose largely from the poor economic performance of the period. Specifically, an adverse economic outlook was the most frequently
cited reason for the credit tightening by banks while reduced financing needs was the
most frequently cited reason for weak loan demand.
The surveys, particularly the SLO, indicate that weak loan growth by small businesses was more often due to credit stringency than to weak loan demand during the period.
For large and medium sized firms, however, the supply restraint may have been most
important in 1990 with subsequent slow loan growth associated with soft demand for
loans.
The survey evidence also suggests that credit tightening had largely stopped and actually began to reverse by 1992 and early 1993. Large business borrowers did not indicate strong borrowing plans, at least compared to mid-1980s levels, suggesting that
weak demand for bank loans was the primary influence on large business borrowing. In
contrast, small business borrowers reported an increased demand for bank loans by 1992
but bank willingness to lend had improved less for them than for large borrowers.
Even so, it is not possible to determine from the narrative analysis whether the supply constraint was unusually strong given the weak economic climate of the period. In
the remainder of the paper, we use statistical models of credit to determine whether the
recent credit tightening was similar to the tightening in earlier business credit crunches.
A Framework for Modelling the Survey Data
The descriptive analysis of the surveys suggests that both demand and supply for business bank loans weakened during 1990-92 relative to the mid-1980s. Although descriptive analysis can identify possible tightness, it cannot easily indicate the relative
importance of supply and demand in generating slow loan growth. In this section, we
outline a model of the market for bank credit and use that view to motivate adjustments
to the survey variables that are designed to measure the noncyclical component of shifts
in loan supply. The following section presents our adjustments and uses them to measure the extent of credit tightness in the recent period.
Our model of the bank credit market is based on a stylized description of credit demand and supply. We assume that a borrower's demand for credit is derived from the
need for external funds — financing needs minus cash flow. In this model, both cash
flow and financing needs depend on the projected demand for the firm's output which
will, in turn, move with the economy-wide level of activity. The cost of loans should
also influence the attractiveness of borrowing. However, a firm is likely to reduce its
borrowing and spending plans if it will not receive the amount of loans that it wishes at
current borrowing rates.20 In aggregate, the demand for loans should increase with the
level of activity while it should decline when either the cost of borrowing or perceived
credit rationing tightens.
Footnote 19 continued
Committee on Banks and New York State Senate Committee on Housing and Community Development,
March 26, 1992, pg. 2.
20
Albert Wojnilower has argued that direct restrictions on credit supply have played a central role in macroeconomic fluctuations in 'The Central Role of Credit Crunches is Recent Financial History," Brookings
Papers on Economic Activity, 1980:2 and "Private Credit Demand, Supply, and Crunches - How Different
in the 1980's?" American Economic Review, Paper and Proceedings, May 1985.
231
Causes and Consequences
232
To model lenders, we follow the literature on lending behavior when borrower quality is observed imperfectly.21 We assume that the market for bank loans is competitive
so that the supply of loans will increase as the lending rate rises relative to the general
level of interest rates, primarily because the cost of raising deposits will likely rise as
more funds are required for lending. The creditworthiness of the pool of potential borrowers should also affect lenders1 supply of loans. More specifically, a deterioration in
creditworthiness should be associated with a reduction in the supply of loans and, consequently, with an increase in the observed loan rate relative to market rates. However,
banks may also react to worsening creditworthiness by adjusting standards or other
terms of lending. Stricter standards or other terms can reduce the fraction of high-risk
potential borrowers in a borrowing pool. For example, high-risk borrowers would be
more reluctant than low-risk borrowers to provide loan collateral that would be forfeited
if the project failed. While raising collateral requirements raises the cost of loans for all
borrowers, the higher probability of failure associated with high-risk borrowers increases the expected value of their loss.
In a recession, both loan demand and supply should shrink. Initially, loan demand
might actually increase as firms borrow to solve what are viewed as temporary cash flow
problems. Eventually, however, as the extent of the downturn in demand becomes evident, the demand for loans drops both because firms attempt to reduce inventory and because fewer investment projects seem profitable. Similarly, loan supply contracts as the
actual or perceived creditworthiness of borrowers drops, signalled by a deterioration in
the quality of loan portfolios.22 As a consequence, loan growth slows as the price and
nonprice terms of loans rise.
There are three survey measures of credit market conditions that are available in reasonably long time series: the net percent of small business borrowers reporting that
loans are harder to obtain, a figure drawn from the NFIB surveys; the net percent of
banks reporting a further tightening of credit standards, from the SLO surveys; and the
net percent of the Fortune 1000 Industrial firms planning to increase their bank borrowing, from the CFO surveys. These measures provide different information about loan
supply and demand and, thus, require somewhat different adjustment to obtain measures
of loan supply shifts.
The NFIB measure of borrower perceptions of credit tightness probably reflects both
supply and demand conditions. Since borrowers are likely to report an increase in credit
tightness whenever a restriction of price or nonprice terms occurs, the NFIB variable
could indicate tightness if either a reduction in loan supply or an increase in loan demand happened.23 For example, as the economy grows, an increase in loan demand that
results in higher price terms is likely to lead to increased reports of difficulty in obtaining credit, as occurred in 1988. However, such reports would also occur if both credit-
21
For an overview of this literature, see D. Jaffee and J. Stiglitz, "Credit Rationing," in B. Friedman and F.
Hahn, The Handbook of Monetary Economics, Vol 2.
22
Ben Bernanke and Cara Lown, in 'The Credit Crunch," Brookings Papers on Economic Activity, 1991:2,
also attempt to isolate unusual movements in loan supply, defining a credit crunch as "a significant leftward shift in the supply for bank loans, holding constant both the safe real interest rate and the quality of
potential borrowers.".
23
The NFIB reports a credit availability measure based on all respondents to its survey. A second credit
availability measure for respondents who borrow regularly (at least once a calendar quarter) can be
derived from the NFIB reports. (See the Appendix for details.) We emphasize the measure using the number of regular borrowers because these respondents sample lending conditions regularly and thus are most
able to determine if credit conditions have changed.
worthiness and economic activity deteriorate, raising nonprice terms of lending. To
purge the NFIB measure of demand-related variation, we removed from the measure
those movements correlated with both general economic activity (business GDP) and
the cost of funds (spread of the prime rate over the six-month CD rate). Since these macroeconomic variables will reflect, to some extent, cyclical movements in supply, our adjusted NFIB measure should not be interpreted as a proxy for all shifts in loan supply.
Rather, the adjusted NFIB series gives an indication of noncyclical movements in supply, including movements previously identified as credit crunches.24 In fact, to the extent
that the spread may be capturing some noncyclical supply shifts, the adjusted measure
may understate noncyclical movements in supply.25
In contrast, the SLO questions on credit standards provide a direct proxy for shifts in
the loan supply.26 Lenders use standards as a screening device that can be adjusted to
achieve a particular level of quality in the potential borrower pool. However, the supply
of loans is likely to vary with the level of economic activity as banks adjust their credit
standards to reflect changes in creditworthiness as well as changes in the cost of funds
to banks. Hence, in our subsequent statistical work, we purged the SLO series of variation related to economic activity (the growth rate of real business GDP) and to the cost
and availability of funds to banks (proxied by the change in the six-month Treasury
rate). This adjusted SLO series may be interpreted as an indicator of noncyclical shifts
in loan supply since it measures loan supply shifts unrelated to changes in activity or the
cost of funds. Again, to the extent that changes in the cost of funds proxy supply shifts
unrelated to cyclical developments, the adjusted SLO measure will understate noncyclical credit supply shifts.
Unlike the other two survey measures, the CFO measure of firm borrowing plans
most likely reflects only loan demand which, in turn, should be closely related to firms'
demand for external funds. Thus, the CFO series for borrowing plans should be closely
related to the components of demand for external funds (growth in cash flow, inventories, and capital spending) and past borrowing (lagged loan growth). We removed from
the CFO series the component related to demand for external funds and to past borrowing, yielding an adjusted series measuring shocks to loan demand. While this adjusted
series partly measures the extent that borrowers are discouraged from borrowing by perceived credit restrictions, it also captures innovations in funding that discourage businesses from bank borrowing.
Although the adjusted CFO measure provides some insights about demand shocks
24
William Dunkelberg has used the survey series directly to detect credit crunches, arguing that a credit
crunch occurs when there is "widespread reluctance of banks to lend to borrowers to whom they would
have lent eagerly under similar economic circumstances" in "Small Business Credit Crunch," The NFIB
Foundation, mimeo, December 1992. This comparison would include in a credit crunch those shifts in
loan supply that accompany deterioration in borrower quality during a recession.
25
Since the spread of the prime rate over the C D rate probably responds to supply shifts, inclusion of the
spread in the regression model removes some part of noncyclical supply shifts from our adjusted NFIB
measure. For example, a restriction in supply, whether arising from deteriorating borrower balance sheets
or simply from bank decisions to reduce lending, will lead to measures that reduce the attractiveness of
bank borrowing, often including a combination of both increases in the interest rate on loans as well as
tighter standards. To the extent that we purge these elements from the NFIB measure, changes in our
adjusted measure should underestimate actual shifts in supply.
26
T h e SLO survey questions about credit standards for business lending have changed over time. Hence w e
allow for differences across survey questions in our empirical analysis. In addition, since the SLO questions on business credit standards were discontinued from 1983 to 1989, we use interpolation procedures,
described in the appendix, to fill in the missing values.
233
Causes and Consequences
and may occasionally reflect perceived credit restrictions, the focus of this analysis is on
the NFIB and the SLO measures. The residuals from our regression models for the NFIB
and SLO survey variables, our adjusted survey measures, indicate loan market tightening beyond that explainable by activity or interest rates. These adjusted survey variables
should not be interpreted as indicators of all supply shifts, since they are constructed by
removing changes attributable to activity or to interest rates—some of which may reflect
supply movements. wSince supply is likely to move with activity (and negatively with interest rates), our adjusted measures are likely to understate the overall size of the supply
contraction accompanying an economic slowdown. However, these adjusted series do
allow us to measure the average tightening in loan availability observed in past recessions and to compare 1990-92 to earlier credit crunch periods, such as 1974 and 197982, to determine if this period was similar to those earlier periods. Of course, this analysis cannot identify the causes of tightening, for which many possibilities exist.
We also developed loan growth models that measured the impact of loan supply after
allowing for the effect of major determinants of loan demand, including demand for
funds (proxied by growth in capital spending and in inventories) as well as the cost of
bank loans (proxied by the spread of the prime rate over the six-month commercial paper rate). We tested whether either the SLO or the NFIB survey measures help predict
loan growth after accounting for economic activity and lending rates. Finding this to be
the case, we use this specification to see whether supply restraint induced particularly
slow loan growth over the past three years compared to both earlier average experience
and to earlier credit crunch periods.
IV. Regression Analysis of Survey Measures of Credit Availability and Loan Demand
Our empirical analysis used regression models to purge survey data of cyclical movements that typically accompany a slowdown in activity. These models allowed us to
measure the size of the recent shift in loan supply. The residuals from our regressions
indicate whether recent supply shifts are atypical, compared both with the last fifteen to
twenty years and with credit crunch periods. Our detailed specification varied by survey
because the NFIB, the SLO, and the CFO surveys have different information about
shifts in loan supply.
Small Business Borrowers: NFIB Surveys
234
The NFIB regular borrower series is an indicator of small business credit conditions.
By this measure, the period containing the second half of 1990 and 1991 exhibited significant tightening of credit, exceeded only by two or three episodes in the past (see Chart 5).
However, as noted earlier, the NFIB series likely measures movements in both loan
supply and loan demand since increases in demand may raise loan rates and, as a result,
cause some respondents to report that credit is harder to obtain. To extract the component of the NFIB series that cannot be attributed to demand shifts, we estimated a regression model relating the series to variables that might influence loan demand. These
explanatory variables include the growth of business GDP, business GDP growth during
recessions (allowing for possible asymmetry in the response of loan demand to activity),
and the spread of the prime rate over the secondary market six-month CD rate (Table
8). 27
We estimated the component of the NFIB series not directly influenced by demand
27
The NFIB reports lhal nearly half of all small business loans are flouting rate loans that are priced off the
prime rale.
using the residuals from our regressions estimated through 1988 and then extrapolated
out of sample (Chart 14). These residuals show that the recent credit episode was at least
comparable to earlier episodes that are viewed as credit crunches. A simple statistical
test of the hypothesis that the mean survey tightness in the recent period was the same
as that observed during the credit crunch episodes in 1975 and 1980-1982 was constructed by rerunning the regression over the entire sample and testing whether dummy
Table 8: Models of Credit Stringency
NFIB Credit Availability Series, Dependent Variables: NFIB Credit Availability; 1973-IV to 1988-IV
Model I
Percent of Regular Borrowing
Model II
Percent of All Respondents
Independent Variables
Coefficients
Coefficients
Constant
30.06(6.71)
12.53(3.21)
t
-.40 (.23)*
-.14 (.10)
t-1
-.33 (.27)
-.11 (.12)
t-2
-.31 (.27)
-.16 (.12)
t-3
-.18 (.25)
-.17 (.11)
t-4
-.01 (.22)
-.01 (.10)
t
-.68 (.50)
-.39 (.22)*
t-1
.13 (.59)
.00 (.26)
t-2
.49 (.61)
.30 (.27)
t-3
-.14 (.58)
.01 (.26)
t-4
-.24 (.46)
-.06 (.21)
Business GDP growth
Business GDP growth when negative
Spread of prime rate over six-month CD rate
2.03(1.39)
1.24 (.63)
t-1
-8.23(1.33)**
-3.84 (.60)**
P
.85 (.08)**
.88 (.08)**
10.93
15.68
.80
.82
t
Ljung-Box Q-Statistic (15 df)
Adj. R2
Note: Independent variables include the annnualized one-quarter growth rate in real business
GDP, the one-quarter real business GDP growth when negative and zero otherwise, and the
quarterly average spread of the prime rate over and the secondary market six-month CD
rate. The dependent variables are the firms reporting credit harder to obtain relative to three
months earlier. In the first model, the dependent variable is the net percent of borrowers; in
the second model, it is the net percent of all respondents. Standard errors are reported in
parentheses. R2 reported for model in p-differences.
* Statistically different from zero at 10 percent level.
** Statistically different from zero at 5 percent level or lower.
235
Causes and Consequences
variables for the credit crunch periods have the same coefficient.28 The data are consistent with the hypothesis that loan supply dropped by the same amount in the three credit
crunch periods.29
Overall, the statistical results for the NFIB consistently indicate that the 1990-92 episode was a period of tight credit supply to small business borrowers. In addition, controlling for the macroeconomic environment suggests that credit tightness was at least
as severe as tightness during earlier credit crunch periods.
Bank Lenders: Senior Loan Officer Survey
The SLO survey gives a fairly direct measure of shifts in the loan supply schedule of
banks, the net fraction of banks tightening standards. Our measure, described in more
detail in the Appendix, was developed from three SLO survey questions. Although there
is a gap in the SLO series from 1983 to 1989, the survey was quite high in 1990 compared to earlier values, suggesting unusual credit tightening (Chart 15).
Although the SLO series provides an indicator of movements in loan supply, we argued above that loan supply would naturally shift as the creditworthiness of borrowers
or as banks' availability of funds changed, suggesting that a measure of supply shifts ex28
Following John Ryding, "Housing Finance and the Transmission of Monetary Policy," Federal Reserve
Bank of New York Quarterly Review, Summer 1990, we identify a credit crunch as a period when the
spread between the Treasury bill rate and Regulation Q ceiling rates is large. He identified the following
periods as crunches: 1969-III tol970-III; 1973-IV to 1975-1; 1979-IV to 1980-III; and 1981-1 to 1982-11.
For this regression we use the 1979-80, 1981-82, and 1990-92 periods to define our credit crunch dummy
variables.
29
Our F-statistic is 0.6, which has a marginal significance level of 62 percent. The same calculation applied
to Model II in Table 8, the "all respondents" specification, gives an F-statistic of 1.4 with a lower marginal
significance level around 25 percent, also consistent with the hypothesis that the episodes are the same.
Chart 14: Adjusted NFIB Credit Stringency Series
236
Net Percent of Borrowers
30
1974
76
78
80
82
84
Source: Residuals from regression in Table 8.
88
90
92 93
eluding these components would be preferable. We used residuals from a regression
model for the survey variable, designed to capture measurable shifts in loan supply, to
measure bank credit tightness beyond that induced by declining creditworthiness or by
restricted deposit funds.
Our model related the credit standard variable to the current value and two lags of
the growth rate of real business GDP—a proxy both for changes in the creditworthiness
of business borrowers as well as a proxy for activity—and to the change in the sixmonth Treasury bill rate—a proxy for banks1 cost and availability of funds. We allowed
the slope coefficients of GDP growth to differ between recessions and other periods; this
specification can detect if GDP changes have a stronger impact on creditworthiness during recessions. Because the survey question on credit standards has been modified over
time, we used interactive dummy variables to permit the regression coefficients to shift
in line with changes in the question.30
The regression model was estimated using quarterly data from 1967 through 1983.
The regression coefficients, shown in Table 9, are consistent with those of a loan supply
schedule; greater GDP growth leads to looser standards, as would be expected if creditworthiness improved with output growth, and an increasing Treasury-bill rate gives rise
to tighter standards. GDP growth affects standards more strongly in recessions, implying greater reductions in supply than would have been obtained using the coefficients
from expansions alone. Standards as measured during 1978-83, when the SLO question
30
The SLO reported changes in standards for new loans through 1977, changes in standards for prime rale
loans during 1978-83, and changes in standards for loans to different size businesses since 1989.
Chart 15: Tightening of Credit Standards
Senior Loan Officer Survey
Net Percent of Banks (Weighted)
100
1967 69
71
75
77
79
81
83
85
87
89
91
93
Source: Board of Governors, Federal Reserve System.
Note: The chart reports the weighted net percent of surveyed banks who for 1967-1977
reported tightening credit standards for loans to nonfinancial businesses; for 1978 to 1983
reported tightening standards to qualify for the prime rate; and for 1990 to 1993 reported
tightening standards to approve C&l loans or credit lines, according to size of firm.
237
Causes and Consequences
Table 9: Model of Credit Stringency
Senior Loan Officer Survey, Dependent Variable: SLO Credit Standards; 1967-1 to 1983-1V
238
Coefficients
Independent variables (1967-83)
Constant
25.29 (6.20)**
Change in T-bill yield
t
14.07(2.50)**
t-1
5.39(2.01)**
t-2
5.83 (2.48)**
Business GDP growth
t
-.13 (.58)
t-1
-1.28 (.53)**
t-2
-.56 (.47)
Business GDP growth during recessions
-1.12(1.07)
Additional variables (1978-83)
Constant
-21.81 (8.94)**
Change in T-bill yield
t
-11.39(3.09)**
t-1
-4.69 (2.82)*
t-2
-7.28 (3.54)**
Business GDP growth
t
.01 (.92)
t-1
1.07 (.86)
t-2
1.02 (.82)
Business GDP growth during recessions
-.62(1.46)
P
.63 (.13)**
Adj. R2
Ljung-Box Q-statistics (17 df)
.65
25.59
Note: Independent variables include the one-quarter change in the one-year T-bill yield; the
annualized one-quarter growth rate in real business GDP; the annualized one-quarter
growth rate in real business GDP during recessions and zero otherwise; and additional variables detecting the change in the coefficients for 1978-83. The dependent variable is the
weighted net percent of banks reporting tighter credit standards in the Senior Loan Officer
Survey. Standard errors are reported in parentheses. R2 reported for model in p differences.
^Statistically different from zero at 10 percent level.
"Statistically different from zero at 5 percent level.
measured the net fraction of banks reporting higher qualifying standards for a prime rate
loan, were largely less sensitive to both activity and interest rates than the more general
question on standards asked earlier.
To measure the component of supply shifts not attributable to changes in borrower
quality (as proxied by GDP growth) and changes in banks1 cost of funds, we computed
residuals from our regression by extrapolating over 1990-92. The residuals, shown in
Chart 16, were derived using standards for loans to both large and small businesses. By
these measures, the recent restriction in supply seems almost as severe as the 1974-75
credit crunch and somewhat more severe than observed over the 1980-82 dual recession.
Large Borrowers: Fortune CFO Surveys
In contrast to the SLO survey, which measures lender perceptions of changes in credit
standards, and the NFIB survey, which measures small borrower perceptions of credit
stringency, the CFO survey provides a measure of loan demand for large corporations.
We used this survey to indicate whether large business loan demand was unusually low
during the recent period, particularly when compared with demand in other periods of
slow activity growth or perceived tight credit. In addition to providing direct information on loan demand, the analysis of the large borrower data offers some indirect evidence on the credit constraints perceived by large firms. By this measure, large and midsized business loan demand did not become consistently weak until midway through
1992, after the recession (see Chart 3).
To measure the extent of unusually slow loan demand in the recent period, we developed a simple model for the net percent of Fortune 1000 industrial firm CFOs that
planned to increase their bank borrowing over the next six months relative to the previ-
Chart16: Adjusted Credit Standards
Senior Loan Officer Survey
Net Percent of Banks (Weighted)
60
Source: Residuals from regression in Table 9.
Note: The survey asked specifically about loans to small firms as of 1990-11, and about
loans to large firms as of 1990-111. Between 1967 and 1983, the survey was about loans to
firms of all size.
239
Causes and Consequences
ous six months. The regression model related the survey variable to macroeconomic
variables likely to influence loan demand. To measure business demand for external financing, we included the six-month real growth rates of cash flow, inventories, and capital spending in the regression. To measure the persistence in bank loan demand, we
included the lagged six-month growth rate of bank loans to nonfinancial corporate businesses. Finally, we allowed the slopes of these variables to shift during recessions, permitting loan demand to respond differently to business activity during those periods.
The regression results indicate that growth in capital spending or inventories raised
demand for loans while growth in cash flow reduced demand, as expected (Table 10).
Lagged loan growth was also significant, suggesting relatively slow adjustment of actual loans to desired borrowing. To purge the survey series of the interest rate effects, we
tried a regression specification (not reported here) that included the change in the spread
of the prime rate over the Treasury rate. Our reported regression omits this variable because it was not significantly different from zero. Moreover, the estimated coefficients
were positive, suggesting reverse causality.
Chart 17 plots the residuals from our regression estimated from 1976 through 1988
and then extrapolated out of sample through 1992. We found that many more firms
planned to increase their lending through the recent period than would have been predicted by the observed slow activity growth. These relatively robust plans could have
arisen if large firms anticipated a quicker recovery in activity than was actually observed
during and after the recession. Further, these large firms apparently did not expect to
face significant credit constraints over the period, since many planned to increase their
future borrowing.31
31
If large borrowers generally expected credit restrictions to be short-lived, this conclusion does not necessarily conflict with our earlier descriptive analysis that suggested that large borrowers perceived some
credit restraint recently.
Chart 17: Adjusted CFO Plans to Increase Bank Borrowings
Goldman Sachs Survey
240
Net Percent of Firms
60
1976
78
80
82
84
Source: Residuals from regression in Table 10.
86
88
90
92
93
To test more formally whether demand was unusual in the recent period, we reestimated the regression over 1976 through 1992, including dummy variables for the credit
crunch periods of 1980, 1982, and the 1990-92. Although the data appear to support the
hypothesis that the mean survey response for this episode was the same as that in the
earlier crunch periods,32 there is strong evidence that the CFO equation shifted after
1988. 33 Thus, the relation between large business borrowing plans and activity indica-
32
The F-statistic for the hypothesis that all three dummy coefficients are equal is 0.37 with a marginal significance level of 69 percent, a result consistent with the hypothesis of equality.
33
When we test the hypothesis of coefficient constancy between the periods 1976-88 and 1976-92, we obtain
an F(8,13) of 2.75 with a marginal significance level of five percent, allowing us to reject the hypothesis
that the regression was stable after 1988 at the five percent level.
Table 10: Model of Bank Loan Demand
Goldman Sachs CFO Survey, Dependent Variable: Fortune 1000 Borrowing Plans; 1976-H1 to 1988-H2
Independent Variables
Coefficients
Constant
-4.17(2.56)
Growth of capital
2.00 (.85)**
Growth of inventories
1.74 (.42)**
Growth of cash flow
-.28 (.15)*
Loan Growth of C&l Loans
t-1
.37 (.20)*
t-2
.34 (.20)
Growth of capital during recessions
3.53(1.42)**
Growth of inventories during recessions
.28 (.84)
Growth of cash flows during recessions
-.12 (.45)
Growth of C&l loans during recessions
t-1
.84 (.48)*
t-2
.73 (.30)*
Ljung-Box Q-statistic (6 df)
Adj. R 2
2.56
.77
Note: Independent variables include the annualized two-quarter growth rates in real capital
spending, real inventories, and in cash flow; the lagged annualized two-quarter growth rate of
real bank loans to all nonfinancial businesses excluding farms, the annualized two-quarter
growth rates in real capital spending during recessions and zero otherwise, the real inventories
during recessions and zero otherwise, and in cash flow during recessions and zero otherwise,
and the lagged annualized two-quarter growth rate of real bank loans to all nonfinancial businesses excluding farms, during recessions, and zero otherwise. The dependent variable is the
net percent of Fortune 1000 firms expecting to increase bank borrowings in the ensuing six
months. Standard errors are reported in parentheses.
'Statistically different from zero at 10 percent level.
"Statistically different from zero at 5 percent level.
241
Causes and Consequences
tors apparently changed after 1988. The large residuals derived from the 1976-88 model
suggest that large business in recent years had more robust borrowing plans—given
growth in funding needs—than would have been typical before 1989.
The results strongly point to the importance of demand factors for the 1990-92 loan
growth at large corporations. Although many large and mid-sized firms planned to reduce their borrowing during 1989-92, the fraction of corporations planning to increase
borrowing was larger than would have been predicted given the sustained slow activity
growth over the period. Since the corporations with relatively strong borrowing plans
likely did not perceive strong credit restraints, this evidence suggests that demand
played a significant role in loan growth weakness at large businesses.
V. Loan Growth and Survey Evidence
Finally, we modelled growth in bank lending to businesses to determine whether the recent weakness in bank lending was primarily attributable to shifts in supply or to declining demand resulting from weak activity growth. We developed models relating loan
growth to standard demand factors and to the survey variables. Since each of our survey
variables gives us some information about supply conditions, our loan models allow us
to determine if the recent supply shifts would have implied particularly slow loan
growth, after accounting for the impact of demand factors.
Senior Loan Officer Survey
242
Our basic model related bank loan growth to the current and four lagged values of capital and inventory spending growth (Table 11). Shifts between bank loans and other
short-term credit, induced by changes in relative price, were modelled by the spread of
the prime rate over the six-month commercial paper rate. Bank loan supply was measured by the SLO credit standards series for small borrowers (plus four lags). As in our
earlier analysis of the SLO survey variable, we modelled the change in the measured
SLO survey question during 1978-1983 by allowing the SLO slope to shift during that
period. We also included lagged loan growth to capture persistence in loan demand.
The model fit is reasonable and the estimated signs are as expected, with higher activity growth inducing greater loan growth, a higher cost of bank loans inducing substitution away from bank loans, and reductions in loan supply (as proxied by increases in
credit standards) lowering loan growth.34 The SLO survey coefficients, as a group, are
significantly different from zero.35 Hence, the SLO survey contains information in addition to the macroeconomic and interest rate variables.
This model allowed us to estimate the component of loan growth attributable to
shifts in loan supply over the historical period. We compared loan growth under the condition of actual credit availability with loan growth under average availability over the
period. First, we estimated the loan model over 1967-88 and then extrapolated out of
sample over the 1989-92 period. The predicted values for loan growth from this regres-
34
We interpret the survey variable as a measure of tightening or easing in bank loan supply. Since we also
include the prime rate relative to a commercial paper rate, the survey coefficients could be interpreted as
the impact of loan supply shifts reflected in other terms or standards of lending. However, some price
effects are probably still captured by the survey variable since the prime rate is not a perfect proxy for
bank lending rates.
35
The observed F-statistic is F(5,56)=4.51 which would reject the null hypothesis that the SLO coefficients
arc jointly zero at the .(X)l level.
Table 11: Loan Growth Model
Senior Loan Officer Survey 1967-1 to 1988-111, Dependent Variable: C&l Loan Growth
Independent Variables
Coefficients
Constant
2.12(4.53)
Growth of capital
t
.43(.42)
t-1
-.14 (.37)
t-2
.25 (.36)
t-3
-.36 (.34)
t-4
.64 (.32)*
Growth of inventories
t
.07 (.22)
t-1
.38 (.24)
t-2
.09 (.23)
t-3
.25 (.23)
t-4
.29 (.21)
Prime six-month CP rate spread
t
-4.44(1.83)**
t-1
.54(1.95)
t-2
1.25(1.92)
t-3
-2.77 (2.09)
t-4
5.82 (2.07)**
SLO survey
t
.10 (.09)
t-1
-.22 (.11)**
t-2
.01 (.11)
t-3
-.17 (.10)*
t-4
-.06 (.08)
Note: Independent variables include annualized one-quarter growth rates in real capital spending growth and in inventory growth, the quarterly average spread of the prime over the sixmonth commercial rate, the weighted net percent of banks reporting tighter credit standards,
the weighted net percent of banks reporting tighter credit standards for loans to qualify for the
prime during 1978 through 1983 and zero otherwise, the lagged dependent variable, and the
annualized one-quarter growth rate in real business GDP during recessions and zero otherwise. The dependent variable is the annualized one-quarter growth rate of bank loans to all
nonfinancial businesses excluding farms. Standard errors are reported in parentheses.
^Statistically different from zero at 10 percent level.
"Statistically different from zero at 5 percent level.
(Continued)
243
Causes and Consequences
Table 11: Loan Growth Model (Continued)
Senior Loan Officer Survey, Dependent Variable: C&l Loan Growth; 1967-1 to 1988-111
Independent Variables
SLO survey during 1978-83
t
-.02 (.20)
t-1
-.31 (.25)
t-2
-.24 (.25)
t-3
.25 (.25)
t-4
-.09 (.22)
Loan growth
t-4
.43 (.10)**
Business GDP growth during recessions
1.42 (.60)**
Ljung-Box Q-statistic (22 df)
Adj. R2
21.94
.55
Note: Independent variables include annualized one-quarter growth rates in real capital
spending growth and in inventory growth, the quarterly average spread of the prime over the
six-month commercial paper rate, the weighted net percent of banks reporting tighter credit
standards, the weighted net percent of banks reporting tighter credit standards for loans to
qualify for the prime during 1978 through 1983 and zero otherwise, the lagged dependent
variable, and the annualized one-quarter growth rate in real business GDP during recessions and zero otherwise. The dependent variable is the annualized one-quarter growth rate
of bank loans to all nonfinancial businesses excluding farms. Standard errors are reported
in parentheses.
^Statistically different from zero at 10 percent level.
"Statistically different from zero at 5 percent level.
sion measure loan growth given actual economic activity and the actual loan supply
stance of banks during the entire 1967-92 period. We also computed a second set of predictions from the loan model, using the mean of the SLO survey variable over noncrunch periods. This second set of predicted values measures loan growth under identical macroeconomic circumstances as the first set, but with our loan supply indicator, the
SLO variable, set to the average non-crunch level during the subperiods.36 Hence, the
difference between the two series measures the impact of loan supply shifts on commercial and industrial loan growth.
Using the mean SLO value during either the 1970s or the 1980s as a reference, we computed two sets of comparisons. Results using either reference period suggest that business
borrowers in the current period experienced a credit crunch comparable to the crunches of
the 1970s and 1980s (Chart 18). If average standards of the 1970s are used as a reference
for the recent period, the drop in loan growth during the 1990s was similar to crunches of
the 1970s, although the recent peak effect of tightening on loan growth never reached 1970s
levels. Alternatively, when we compare the current period with credit standards of the
1980s, the recent tightening in loan supply is similar to the crunch in early 1980.
36 The
244
periods arc: 1969-III to 1970-III, 1973-IV lo 1975-1, 1979-IV to 1980-111, and 1981-1 to 1982-11.
Chart 18: Effect of Credit Stringency on C&l Loan Growth
Senior Loan Officer Survey
1967 69
71
73
75
77
79
81
83
85
87
89
91
93
Source: Derived from the loan growth model in Table 11.
Note: Shown are the differences between predicted loan growth using the actual survey and
predicted loan growth using non-crunch.means. A negative value denotes where predicted
loan growth was lower due to tighter than predicted credit standards.
NFIB Survey
To create a comparison with the SLO results, we modified our basic loan model by adding the NFIB series measuring the net fraction of small business borrowers who find
credit harder to get. For the loan analysis, we used the adjusted NFIB survey measure
presented earlier; this adjusted measure may be a better proxy for loan supply shifts because it attempts to purge the NFIB series of reported tightening that could be associated with shifts in loan demand. The regression signs seem generally plausible and the
NFIB survey appears to contain information beyond the demand determinants that can
explain loan growth (Table 12).37
To measure the strength of loan restrictions over the 1990-92 period, we compared
predicted loan growth derived using the realized NFIB values in the recent period with
predictions using average survey values in earlier non-crunch periods. The difference
between these two measures is a proxy for the restriction in loan growth induced by unusually tight loan supply.38 This comparison suggests that loan supply induced unusually weak 1989-92 loan growth compared with the historical experience, including
earlier credit crunches (Chart 19). Thus, the use of the adjusted NFIB series suggests a
significant recent tightening in loan supply, a finding that is consistent with the SLO
37
The F-statistic for the hypothesis that all the NFIB coefficients are zero was F(5,61) = 2.67 which implies
that the hypothesis would be rejected using a 3 percent significance level.
38
The chart is virtually identical whether we use means for the 1970s or the 1980s for the out-of-sample
extrapolation.
245
Causes and Consequences
Table 12: Loan Growth Model,
NFIB Survey, Dependent Variable: C&l Loan Growth; 1973-1V to 1988-1V
246
Independent variables
Constant
Coefficient
-15.04(11.51)
Growth of capital
1.44 (.59)**
t-1
.18 (.48)
t-2
.47 (.48)
t-3
-.87 (.51)*
t-4
-.29 (.49)
Growth of inventories
t
.11 (.37)
t-1
.55 (.40)
t-2
-.15 (.35)
t-3
.04 (.32)
t-4
-.09 (.28)
Prime six-month CP rate spread
t
-.51 (2.64)
t-1
2.10(2.90)
t-2
4.72 (2.59)*
t-3
-1.35(2.86)
t-4
3.28 (2.75)
NFIB survey residuals
t
-.10 (.36)
t-1
.24 (.42)
t-2
-.05 (.38)
t-3
-.55 (.37)
t-4
.24 (.37)
Note: Independent variables include annualized one-quarter growth rates in real capital
spending growth and in inventory growth, the quarterly average spread of the prime over the
six-month commercial paper rate, the net percent of firms reporting tighter credit availability,
purged of cyclical economic and interest rate behavior, the lagged dependent variable, and
the annualized one-quarter growth rate in real business GDP during recessions and zero otherwise. The dependent variable is the annualized one-quarter real growth rate of bank loans
to all nonfinancial businesses excluding farms. Standard errors are reported in parentheses.
* Statistically different from zero at 10 percent level.
** Statistically different from zero at 5 percent level.
(Continued)
Table 12: Loan Growth Model (Continued)
NFIB Survey, Dependent Variable: C&l Loan Growth; 1973-1V to 1988-1V
Independent variables
Coefficient
Loan growth
t-4
Business GDP growth during recessions
Ljung-Box Q-statistic (15 df)
Adj. R2
.30 (.14)**
2.95 (.81)**
10.56
.38
Note: Independent variables include annualized one-quarter growth rates in real capital
spending growth and in inventory growth, the quarterly average spread of the prime over the
six-month commercial paper rate, the net percent of firms reporting tighter credit availability,
purged of cyclical economic and interest rate behavior, the lagged dependent variable, and the
annualized one-quarter growth rate in real business GDP during recessions and zero otherwise. The dependent variable is the annualized one-quarter real growth rate of bank loans to
all nonfinancial businesses excluding farms. Standard errors are reported in parentheses.
* Statistically different from zero at 10 percent level.
** Statistically different from zero at 5 percent level.
survey results.39
Finally, we compare the effects of the SLO and NFIB surveys on loan growth (Chart
20). These surveys are complementary in that they represent different perceptions of
loan supply shifts, one from the lender's perspective and the other from the borrower's
perspective. The results broadly suggest that the reduction in loan supply during 19891990 led to a drop in lending growth similar to that observed in earlier credit crunch
periods.
VI. Conclusions
Our paper analyzes survey data to discover the extent of evidence for a credit crunch in
the recent period. There are two major parts to the analysis: The first is a narrative discussion of the surveys. The second is regression modelling of three survey variables and
of business loan growth.
Our narrative analysis suggests that large firms were subject to some credit tighten-
39
We also estimated a loan growth model using the actual NFIB series. When we repeated our procedure
using this model, we found that the recent period did not exhibit as sharp a tightening in loan growth as did
earlier credit crunches. One explanation of the difference in loan growth across the two NFIB series is that
the adjusted NFIB series is a closer proxy for movements in loan supply. For example, the actual NFIB
series suggests that significant shortfalls in loan growth often occur before recessions, perhaps because the
survey reflects perceived tighter conditions arising from the increase in loan rates accompanying rising
loan demand. In contrast, the adjusted NFIB series indicates that loan supply typically declines steadily in
the early stages of recessions.
A second possible explanation for the difference between the two results is a shift in the relative importance of non-price and price rationing of loans. Our adjustment of the NFIB series may remove not only
changes in demand but also the component of tightening typically attributable to price terms of credit. The
results using the original NFIB series presume that the relative importance of price and non-price rationing
is constant over time. The difference between specifications suggests that the non-price channel may have
been used more aggressively during the current episode than in earlier periods. Possible reasons for a shift
toward direct credit rationing would include deteriorating loan portfolios, which might have made default
risk more important than in earlier periods, and capital constraints on banks.
247
Causes and Consequences
Chart 19: Effect of Credit Stringency on C&l Loan Growth
NFIB Survey
1975
83
81
77
85
87
89
91
93
Source: Derived from the loan growth models in Table 12.
Note: Shown are the differences between predicted loan growth using the actual survey and
predicted loan growth using non-crunch means. A negative value denotes where predicted
loan growth was lower due to tighter than predicted credit standards.
Chart 20: Effect of Credit Stringency on C&l Loan Growth
248
Percent
30
20
10
-10
-20
-30
1967 69
71
73
75
77
79
81
83
85
87
89
91
Source: Derived from the loan growth models in Tables 11 and12.
Note: Shown are the differences between predicted loan growth using the actual survey and
predicted loan growth using non-crunch means. A negative value denotes where predicted
loan growth was lower due to tighter than predicted credit standards.
ing starting in 1990 and continuing through early 1993. However, a substantial part of
the weakness in large firm loan growth seems to have arisen from weak demand, a condition that has apparently continued into the present.
In contrast, small firms were subject to tightening starting in 1989. This restriction
in credit continued through 1991, only easing in 1992. Although there is some evidence
of a slowing in small business loan demand, the surveys suggest that weak loan demand
only became an important factor in 1992. Further, small business loan demand appears
to have recovered by 1993.
Our regression models attempted to purge the survey data of both demand movements and cyclical movements in loan supply, yielding proxies for the noncyclical
movements in supply. The adjusted NFIB series implies that credit tightening began
sometime in 1988, became extremely severe by 1991, and has only recently seemed to
slacken. The adjusted SLO series suggests that large firms perceived the most severe increase in credit tightening during 1990 and that the atypical tightening largely began to
reverse in 1991. The adjusted CFO series suggests that expected future loan demand at
large firms was consistently and surprisingly strong in 1991 given the observed weakness in actual demand for funds reported by the surveyed firms.
Finally, models of loan growth were developed using both the SLO and the NFIB
survey measures. Both models imply that tight loan supply led to significantly slower
loan growth in recent years. The model using the SLO series indicates that loan growth
was severely restrained in 1990 and 1991. This result suggests that supply influenced
large firm loan growth most strongly in 1990-91. In contrast, the model using the NFIB
series implies that restraint on loan growth began in 1989 and worsened in 1991 -92. The
supply restraint on small firm lending during the recent period seems to have been longer lasting than the restraint on large firms.
Overall, the survey evidence suggests that the supply of bank credit to business was
tight over 1989-92. The constriction of credit supply to small business was particularly
severe, lasting relatively longer than the restraint on large business.
Appendix
I. The Federal Reserve Senior Loan Officer Opinion Survey
The Federal Reserve System began conducting the Senior Loan Officer Survey on a
quarterly basis in 1964. Since the earliest surveys were considered experimental, summary statistics are available only starting in 1967.4()The information in the survey is primarily qualitative and reflects the judgement of a senior loan officer at each participating
bank regarding that bank's policies governing loans. When policies are changed, the respondent is asked to indicate the nature and degree of change. For example, the officer
indicates whether lending policy has become "moderately" or "much" firmer or easier.
From 1967 through 1977, the survey contained a consistent set of twenty-two ques-
40
The purpose of the survey was to determine whether non-price lending policies of banks became more
restrictive, more lenient, or remained essentially unchanged relative to three months earlier. This focus
reflected the thinking that banks often initially react to changes in the cost and availability of loanable
funds by making marginal changes in their non-price terms and conditions of lending. Hence, such information on a timely basis could "throw additional light on bank responses to changes in monetary policy
and thus be helpful in the formulation of monetary policy."
249
Causes and Consequences
250
tions, including one on banks' credit standards for business loans. The questions were
posed to a sample of 121 banks chosen from the panel of banks already participating in
the Federal Reserve's Survey of Terms of Bank Lending. As such, they were among the
largest banks operating in the national market for business loans, accounting for about
60 percent of outstanding loans.
Since 1977, the survey's format has been revised several times. In February 1978,
some questions, including the credit standards question, were modified to capture bank
interest rate policies and the willingness to make loans at different maturities. In May
1981, the panel was reduced to 60 banks, typically the largest banks in each Federal Reserve District. At that time, the common set of questions was reduced from twenty-two
to six. In 1984, the survey became less consistent. The credit standards question, among
others, was dropped, and the number and types of questions differed from survey to survey. Even the number of surveys varied from four to six per year. Only the question on
the willingness to make consumer installment loans was asked consistently.
The format of the survey has been more consistent since 1990, when questions regarding banks' credit standards for loans to businesses were reintroduced. On a quarterly
basis, banks have been asked about changes in their standards for loans to small, middle
market, and large firms, with separate answers elicited for each group.41 Banks have also
consistently reported the reasons for changing standards and the terms that have
changed. Some of the recent surveys have also contained questions on banks' perceptions of loan demand. In our analysis, we present the weighted net percent of banks responding to each question. For example, our SLO series on tighter standards is
computed by summing the responses after converting to a numerical scale that assigns
a weight of 2 to banks reporting "much firmer" standards, I for "somewhat firmer" standards, -1 for "somewhat easier" standards, and -2 for "much easier" standards.
In its current formulation, the question on banks' credit standards differs from the
questions posed in earlier periods. Since 1990 banks have been asked about "standards
to approve commercial and industrial loans or credit lines" to firms according to size.
By contrast, from 1978 to 1983, banks were asked about "standards to qualify for the
prime rate," and between 1967 and 1977, about "standards for loans to nonfinancial
businesses." Since the question asked between 1978 and 1983 was considerably more
narrow than the earlier, more general standards question, an adjustment in the SLO survey regression models for 1978 to 1983 may be required. We control for the change in
question format between 1978 and 1983 with interactive dummy variables for all the
variables in the regressions. Our survey model, presented in Table 9, supports the need
for a correction since the interactive variables are jointly significant with a marginal significance level of .43 percent.
Furthermore, to interpolate the series for 1984 to 1989, the period during which the
credit standards question was not posed, we used a simple regression model.42 We regress the available SLO series on the change in the one-year Treasury bill yield (plus
two lags), the real business GDP growth rate (plus two lags), the SLO series on the willingness to make consumer installment loans (available consistently from 1967 to the
present) and dummies for all the variables plus the constant term for the 1978-83 period
41
Middle market firms are those with annual sales of between $50 million and $250 million, or as defined by
the institution.
42
This model differs from our SLO model presented in Table 10. Since the consumer installment question is
the only series available throughout the history of the survey, we use it as a proxy for banks' overall willingness to make loans during the period. However, because this scries does not explain commercial and
industrial loan stringency, we omit it from our base model.
(Table Al). We use the predicted values from this regression to fill in the original series
for the 1984-89 period. The regression coefficients all have reasonable values with the
signs predicted by theory. The adjusted R-squared for the interpolation model is 0.59.
Survey of Fortune Chief Financial Officers
Goldman, Sachs & Co. has been surveying chief financial officers from the Fortune
1000 industrial companies semiannually since April 1976. In 1985, 350 Fortune service
companies were added to the potential pool of respondents.
This survey is intended to measure the financial officers' expectations, not firm facts
or plans, regarding their own corporation's bank credit financing trends in the upcoming
half-year relative to the previous half-year. (This approach, Goldman Sachs claims,
avoids problems of seasonality.) The format of the survey is multiple choice to ensure a
high response rate. In the last several years, questionnaires have been mailed to 600 to
850 chief financial officers; responses received within twelve business days of the mailing are included in the published results. The response rate is about 30 percent, though
it has ranged from 26 to 56 percent.
A question on expected borrowing needs has been asked since the inception of the
survey. Goldman Sachs analysts claim that the net percent of firms planning to increase
bank borrowings, computed as the difference between the percentage of firms planning
to increase borrowings and the percentage of firms planning to decrease borrowings, is
a good predictor of the direction of the next half-year's bank business loan growth: twenty-eight of the thirty-two surveys provided correct predictions, even though one of the
misses was in the current episode.43 Consistent with our interpretation of the other survey data, however, we use this series as an indicator of current bank borrowing demand
rather than as a predictor of future loan demand. As Chart A1 shows, the series is more
closely related to current loan growth than future loan growth.
We construct a second indicator of current loan demand from several CFO survey
series (Table A2). Beginning in 1983, the survey queried respondents about their projected inventory growth and capital spending growth. In 1985 a question about internal
cash flow projections was added. We compute a proxy for the net demand for external
financing by subtracting internal cash flow growth from projected inventory and capital
spending growth. Whereas the question regarding intended bank borrowings directly
asks firms about their bank borrowing plans, our proxy provides an indication of firms'
need for external funds. If respondents are consistent in their responses, this measure of
net demand for external funding should be positively correlated with the net percent expecting to increase bank borrowing, as evident in Chart 3. 44
From October 1990 through October 1992, the survey also included questions about
credit availability (Table 5). The questions varied in detail between surveys, and were
omitted altogether after the October 1992 survey.
43
"The
1984,
1989,
1991,
vey.
44
Since the mid-1980s, the borrowing plans measure and the net demand proxy gave contradictory signals
about future loan growth in exactly the two half-years when borrowing plans failed to predict the direction
of actual commercial and industrial loan growth. This result indicates that the net demand for external
financing contains additional information useful in explaining bank loan growth.
four major directional discrepancies were in 1980, following the imposition of credit controls; in
when merger and acquisition financing overtook the cyclical features of borrowing needs; in late
when the Highly-Leveraged Transactions (HLT) business came to an abrupt halt; and Spring of
when stronger intentions evaporated in the face of a continuing credit squeeze." p. 5, April 1992 Sur-
251
Causes and Consequences
Table A1: Model to Interpolate Credit Stringency
Senior Loan Officer Survey, Dependent Variable: SLO Credit Standards; 1967-1 to 1992-1V
252
Independent Variables (1967-92)
Coefficients
Constant
31.51(3.09)**
Change in T-bill yield
t
9.52 (2.95)**
t-1
4.77(2.73)*
t-2
2.53(2.55)
Business GDP growth
t
-0.17(0.66)
t-1
-1.24(0.56)**
t-2
-0.36(0.55)
Business GDP growth during recessions
-2.57(1.30)*
Willingness to make consumer installment loans
0.47(0.17)**
Additional Independent variable* (1970^3)
Constant
-26.91(5.03)**
Change in T-bill yield
t
-16.05(4.80)**
t-1
-12.43(4.54)**
t-2
-8.30(4.12)**
Business GDP growth
t
.64(1.16)
t-1
2.26(.97)**
t-2
1.35(.96)
Business GDP growth during recessions
Willingness to make consumer installment loans
Ljung-Box Q-statistic (26 df)
Adj. R2
-5.22 (2.30)**
0.16 (.32)
31.04
.59
Note: Independent variables include the one-quarter change in the one-year T-bill yield; the
annualized one-quarter growth rate in real business GDP; the annualized one-quarter
growth rate in real GDP during recessions and zero otherwise; the weighted net percent of
banks reporting a decreased willingness to make consumer installment loans in the Senior
Loan Officer Survey; and additional variables detecting the change in the coefficients 197883. The dependent variable is the weighted net percent of banks reporting tighter credit standards in the Senior Loan Officer Survey. Standard errors are reported in parentheses.
'Statistically different from zero at 10 percent level.
"Statistically different from zero at 5 percent level.
The National Federation of Independent Business Survey
The National Federation of Independent Business is a small-business organization with
more than half a million members. This segment of the business community employs
about half of the private, non-farm workforce and produces about half of the nation's
gross private product. By the organization's statistics, about one in every eight employers in the United States is an NFIB member.45
Since 1973, NFIB has surveyed its members on current and anticipated credit and
business conditions. These surveys enable us to gauge small firms' loan demand and perceptions of credit stringency. For the last several years, NFIB has received about 2,000
to 2,200 responses from the 7,000 questionnaires mailed out to a randomly selected
group of members—a response rate of around 30 percent.
Business condition variables reported by NFIB include the net percent of firms reporting higher sales, undertaking capital expenditures, and adding to inventories. These
variables may be indicative of small business loan demand since stronger current or anticipated business conditions encourage businesses to add to inventories and undertake
capital expenditures. Even if earnings were growing, this spending would probably be
financed, at least in part, with credit.
NFIB also reports several credit condition variables: the percent of firms currently
paying higher interest rates than three months earlier, the average interest rate paid by
the reporting sample, the net percent of firms reporting loans currently harder to obtain
45
William C. Dunkleberg, 'The Credit Crunch - Myth or Mistaken Monetary Policy?" NFIB, Oct. 1991, pg. 2.
Chart A1: Goldman Sachs Survey of Fortune Chief Financial Officers
and Business Loan Growth
Percent
20
Net Percent of Firms
40
CFO Survey
(left scale)
15
10
Business loan growth
(right scale)
1976
78
88
90
92 93
Source: Goldman, Sachs and Co., Flow of Funds, Bureau of Economic Analysis.
Note: Shown are the net percent of firms planning to increase bank borrowings in the six
months following the period shown, and the one-year growth rate of bank loans to nonfinancial businesses. The correlation between the survey and current loan growth is .39, and
between the survey and forecasted loan growth is .25.
253
Causes and
Consequences
than three months earlier,46 the net percent of firms expecting loans to be harder to obtain
over the next three months than currently, the percent of firms reporting interest rates
and financing as their single most important problem, and the percent of firms borrowing at least once each quarter~the "regular borrowers". These credit condition variables
46
It is difficult to interpret this series. The question that is asked is, "Arc these loans easier or harder to get
than they were three months ago?" If the same firms repeatedly responded "harder," then the interpretation
is that credit stringency was increasing cumulatively. However, if different firms answer "harder" in different periods, then the availability is worsening in the sense that more firms are being affected, but not in
the sense that the same loans arc getting harder and harder to obtain. Because the sample is not consistent,
we tend toward the latter interpretation.
Table A2: Loan Demand Proxy for Large and Middle Market Firms
254
Inventory
Growth
Capital Spending Growth
Internal Cash
Flow Growth
Net Demanda
1985
H1
H2
1.3
0.6
2.6
1.0
7.4
8.3
(3.5)
(6.7)
1986
H1
H2
1.1
0.7
1.2
0.7
8.0
8.3
(5.7)
(6.9)
1987
H1
H2
1.7
1.9
3.1
2.3
8.5
6.4
(3.7)
(2.2)
1988
H1
H2
1.8
1.1
CO CO
CO CO
Goldman Sachs CFO Survey Projected Growth
8.5
7.1
(2.8)
(2.7)
1989
H1
H2
2.2
1.4
5.6
4.0
8.7
6.4
(0.9)
(1.0)
1990
H1
H2
1.7
0.2
3.5
1.0
7.2
5.1
(2.0)
(3.9)
1991
H1
H2
(0.1)
1.4
0.1
3.5
5.1
9.5
(5.1)
(4.6)
1992
H1
H2
0.3
1.0
2.6
3.2
8.5
8.8
(5.6)
(4.6)
1993
H1
H2
(0.2)
0.7
3.4
3.6
9.9
9.5
(6.7)
(5.2)
Survey Date
Source: Goldman, Sachs, & Co.
Note: Survey is conducted semiannually to project growth for the six months following the
period shown.
a
- Capital spending plus inventory growth minus internal cash flow growth, all projected over the
next six months.
provide several views of borrowers1 perceptions of loan supply conditions. Since small
businesses generally rely heavily on bank credit, their perceptions of credit conditions
most likely reflect bank loan availability. However, perceptions of credit stringency are
likely to be influenced by both higher rates as well as decreased credit availability. Such
perceptions of stringency may lead to expectations that future loans will be harder to obtain as well as reports that financing is an important problem for firms.
Our analysis of small business perceptions of credit availability in the text uses survey responses to the NFIB question whether loans are harder to obtain. Our measures
are derived from the responses to two questions. NFIB first asks respondents:
"If you borrow money regularly (at least once a quarter) as part of your business activity, how does the rate of interest payable on your most recent loan compare with that paid three months ago?"
Then, as a followup to the above question, NFIB asks:
"Are these loans easier or harder to get than they were three months ago?"
The net number of firms reporting credit harder to obtain, calculated as the number
of firms reporting credit harder to obtain minus the number of firms reporting credit easier to obtain, can be normalized by two measures of the number of survey respondents.47
One measure, the dashed-line series in Chart A2, is the net number of firms reporting
that credit is harder to obtain as a percent of all respondents for that survey. The other
series, the solid-line series, is the net number of firms reporting credit harder to obtain
47
In our narrative analysis, we present a series from which survey participants who report credit stringency
but are not regular borrowers are dropped. Our regression analysis, however, uses a scries on which such
an adjustment is not made because, until recently, only unfiltered data were available as a consistent time
series. Although the percent of regular borrowers has declined significantly over the life of the survey, the
difference between the unfiltcred and filtered series is consistent and very small. Thus, our use of the unfiltercd series creates, at worst, only a small bias.
Chart A2: Credit Stringency to Small Firms
NFIB Survey
Percent
60
Net percent
of borrowers
Net percent
of all respondents
1974
76
78
80
82
84
86
90
92 93
Source: National Federation of Independent Business
255
Causes and Consequences
as a percent of the regular borrowers - those firms answering both questions. The two
series differ because the percent of NFIB respondents borrowing regularly has been declining since 1980 (Chart A3). Arguably, declining borrowing activity could be either a
demand- or a supply-induced phenomenon. For example, in examining the "all respondents" series, Dunkelberg finds no evidence of a credit crunch, arguing that declining
borrowing participation is a demand induced phenomenon.
The "regular borrowers" series may be a more appropriate measure for a definitional
reason that is independent of the demand-supply controversy. Using regular borrowers
in the denominator may be more accurate because they are regular participants in the
loan market, and are thus better able to determine whether conditions have changed.
III. NAM Small Manufacturers Operating Survey
The National Association of Manufacturers (NAM) is a national business association
with more than 12,500 member companies and subsidiaries. The firms vary in size,
ranging from very large companies to 9,000 smaller manufacturing firms, with fewer
than 500 employees. According to the association, NAM member companies employ 85
percent of all workers in manufacturing and produce more than 80 percent of the nation's manufactured goods. Typically, about 2,300 to 2,500 questionnaires are mailed in
March of each year, and the response rate is about 26 percent.
NAM surveys its small manufacturer members once a year in the spring. Respondents are asked about current business conditions and their expectations for the upcoming year. Respondents are also asked whether financing and credit are easier to obtain
than they were twelve months earlier, and how their ratio of debt-to-cash flow has
changed in the past year.
Chart A3: Regular Borrowers
NFIB Survey
256
Percent of All Respondents
55
1974
76
80
82
84
86
Source: National Federation of Independent Business
Note: Respondents borrowing at least once every three moths.
88
90
92 93
Because the questions are posed in the first quarter of each year, it difficult to apply
the results to any particular calendar year. We follow the convention that the survey data
reflect economic conditions during the prior calendar year. For example, the 1988 survey mainly reflects credit conditions during 1987.
257
Causes and Consequences
258
Influence of the Credit Crunch on Aggregate
Demand and Implications for Monetary Policy
by Patricia C. Massed
This paper investigates the impact of credit restraint on real activity in the 1990-91 economic downturn. In addition, it examines whether financial system fragility, combined
with institutional changes over the 1980s, made it more difficult for monetary policy to
ease credit conditions and stimulate real activity. These two issues are related because
many of the changes in financial structure and regulation that differentiate this credit
crunch from previous episodes have also altered the transmission mechanisms of monetary policy.
More specifically, this investigation examines whether credit restraint, measured in
several different ways, preceded or caused the 1990-91 recession and the slow-growth
recovery that followed. The paper also documents long-run changes in financial structure and nonmonetary factors that made the cycle unique and contributed to slow
growth. Finally, the paper asks whether traditional relationships between real activity
and economic fundamentals (including monetary policy) accurately predicted the slow
economic growth from 1989 to 1992, or whether proxies for credit restraint are necessary to explain the weakness in activity.
Many of the issues dealt with in this paper relate to whether monetary policy affects
real activity more through a "money" channel or through a "credit" channel.2 Roughly
speaking, monetary policy affects the real economy via the money channel if policy
works exclusively by changing the relative price of money, the interest rate. The credit
channel view suggests that monetary policy also affects the real economy by changing
the quantity and composition of credit, for example by changing the supply of intermediated credit relative to market credit. Because this paper looks at the effects on aggregate demand of both interest rates and shifts in credit supply and demand, the results
described below should help clarify whether monetary policy channels via interest rates
1
My thanks to Charles Stcindcl, M.A. Akhtar, Cara Lown and Ethan Harris for comments on earlier drafts
of this paper. Thanks also to Cynthia Silvcrio and Joshua Glcason for excellent research assistance.
2
See Bernankc (1993) for a discussion of the money and credit channels of monetary policy.
259
Causes and Consequences
or credit behaved in normal ways during the late 1980s and early 1990s. The main conclusions of the study are as follows:
• While credit supply problems probably had detrimental effects on the real economy
after 1989, other factors appear to have contributed to overall slow growth as well.
These include balance sheet restructuring by both firms and consumers, relatively
tight fiscal policy, and extreme pessimism, particularly on the part of consumers.
• Despite worries that a fragile financial system blocked the effects of easier monetary
policy to the real economy, policy was successful in stimulating some sectors of the
economy. With the exception of large-scale real estate, sectors most sensitive to
policy did slightly better, relative to cyclical norms, than the rest of the economy.
Nonetheless, monetary policy did not have the same widespread stimulative effect
on the economy that it had had in the past.
• Both informal observations and econometric estimates suggest a general malaise in
economic activity, particularly in the household sector, that was more widespread
than the credit slowdown and that was not well described by economic fundamentals, including the stance of monetary policy.
The first section of the paper examines patterns in aggregate demand and relates
them to credit and monetary policy, comparing behavior across business cycles. In addition it discusses several proxies for credit restraint in the early 1990s episode and
looks at the timing of changes in credit and demand. Section II discusses several other
possible explanations for the economy's sluggish response to easing monetary policy in
1991 and 1992. Section III presents econometric evidence on aggregate demand and
policy during the early 1990s and relates this evidence to several proxies measuring the
severity of the credit slowdown.
I. Aggregate Demand and Credit
260
From an aggregate demand perspective, the early 1990s "credit crunch" was quite different from historical episodes. The aggregate demand boom preceded the recession by
more than two years, and the slowdown in credit led rather than lagged the slowdown
in activity. Tight monetary policy and disintermediation at depository institutions,
which typified previous credit crunches, were almost completely absent. Instead, the
1989-92 credit slowdown was characterized by subpar growth (interrupted by a short recession), easing monetary policy and falling interest rates, and financial distress at many
intermediaries. Only in the last of these features did it resemble previous crunches.
Many of the factors that made the credit slowdown unique were direct or indirect
consequences of the 1980s "bubble" in debt formation relative to economic activity
(Chart 1). In particular, high leverage contributed to the precarious financial health of
both borrowers and lenders in the late 1980s, which in turn, appears to have been a major
factor in both the slowdown in credit and in aggregate demand.
A number of studies have examined the credit slowdown. For example, see Bernanke and Lown (1991), Johnson (1991), Peek and Rosengren (1991), Hancock and
Wilcox (1992) and in this volume, studies by Cantor and Rodrigues, Lown and Wenninger, Mosser and Steindel. The general conclusion of these studies is that credit demand factors, in particular aggregate demand components and interest rates, could not,
by themselves, explain a significant portion of the 1989-92 credit slowdown.
This section of the paper uses results of these studies to look at the flip side of the
coin: the implications of credit restraint for economic activity. It begins by describing
the behavior of activity and debt from 1989 to 1992. In addition, because aggregate output and aggregate credit are simultaneously determined, this paper (like those listed
above) attempts to separate credit supply and demand shocks in several imperfect, but
hopefully plausible ways. Below, aggregate demand weakness is related to several
proxies that measure the degree of credit supply restraint, including proxies derived
from studies elsewhere in this volume. In addition, the issue of "causality" or timing of
credit relative to aggregate demand is investigated using reduced form relationships between credit aggregates, proxies, and real activity.
A. Cyclical Comparison of Aggregate Demand
The 1990-91 recession was a short, slightly worse-than-average downturn. Chart 2
shows that the economy was quite weak going into the recession, however, and it was
well below par coming out. The entire path of the cycle was a slow roll, up and down
and up again, rather than the typical boom—bust—boom pattern.
One explanation for this atypical cyclical behavior (discussed in detail below) is that
the slowdown in credit growth, in particular restrictions on credit supply by intermediaries, inhibited economic activity and blocked the effects of easing monetary policy.
Under such a scenario, the sluggishness in the economy would presumably have been
concentrated in those sectors that are most sensitive to monetary policy and most closely
connected with credit formation: consumer durables, housing, business investment, and
perhaps net exports. Chart 3 presents these components of real output indexed to business cycle peaks and troughs. Somewhat surprisingly, the recession in these sectors
looked quite average, certainly less severe than the 1981-82 cycle, when monetary policy was substantially tighter. In contrast, growth before the recession was weak, and the
Chart 1: Real GDP and Real Total Nonfinancial Private Credit
Four-Quarter Growth Rate
Real Total Nonfinanclal Private Credit
195860
62
64
66
68
70
72
74
76
78
80
82
84
86
88
90
92
Note: Shaded areas represent recessions.
261
Causes and Consequences
recovery was well below par in spite of steady easing of monetary policy and falling interest rates.
For comparison, sectors less directly sensitive to monetary policy (and generally less
closely connected to credit) are shown in Chart 4. Here both the 1990-91 recession and
the recovery look exceptionally weak. The peak comparison shows slow growth before
the recession and a worse-than-average downturn during the recession—on a par with
the severe 1981 -82 recession. Further, these sectors had almost no recovery in 1991 and
1992.
The widespread weakness in activity can also be seen in aggregate demand components. For example, all categories of consumer spending were weak from 1989 to 1992
(Chart 5). Consumer spending on durables, which is more sensitive to credit and mon-
Chart2: Real GDP
262
Peak = 100
IU4
Average ^r
102
1 ,.. aMI «irtr
i
%iBll
100
90-91
98
/
81-82
/
1
-4
J
•
I
|
-3
-2
-1
-4
-3
1
I
2
3
Quarters
i
i
4
5
6
7
2
3
4
5
Trough =100
112
-6
-5
-2
-1
T
1
Quarters
Note: AVG includes 1960-61, 1969-70, 1973-75, 1981-82 recessions.
i
etary policy, suffered a fairly typical decline during the recession but had a weak recovery thereafter in spite of falling interest rates. Expenditures on nondurables and services
were even weaker in comparison with historical norms. They were a drag on the economy before the recession, and they actually declined in 1990. Their recovery was only
2.2 percent by the end of 1992, compared with more typical 6 percent increases.3
Cyclical comparisons for the investment components are shown in Charts 6 and 7.
Real equipment investment (Chart 6A) was well above previous norms during both the
3
Also sec Blanchard (1993), Lccper (1992), Stcindcl (1992) and Throop (1991) for further discussion of the
unusual weakness in consumer spending both during and after the 1990-91 recession.
Chart 3: Credit and Policy Sensitive Components of Real GDP
Peak = 100
100
90-91
>
95
Average
\
90
/81-82
1
1
1
i
i
1
1
-
1
105 -
100 -
95
u
-6
-5
-4
-3
-2
-1
Note: Includes consumer durables expenditure, residential investment, nonresidential
investment, and net exports.
AVG includes 1960-61, 1969-70, 1973-75,1981-82 recessions.
263
Causes and Consequences
recession and the recovery because of large relative price declines for computers. The
inventory cycle before and during this recession was quite small by historical standards
(Chart 6B), largely because of improved inventory management techniques adopted
over the 1980s. Nonresidential construction, pictured in Chart 7A, was incredibly weak
during the latest cycle. Here the low level of activity appears to have been the result of
massive overbuilding in the mid 1980s and perhaps credit constraints as well. Whatever
the cause of this weakness, easier monetary policy and falling interest rates had little or
no short-run effect on activity during the recovery. Housing (Chart 7B) had a fairly typical decline duriilg 1990 but recovered more slowly and did not seem to respond quite
as much to monetary easing as it had in the past. However, long-run changes in financial
Chart 4: Components of Real GDP Less Sensitive to Credit and Monetary Policy
264
Peak = 100
106
104 -
-4
-3
-2
-1
P
1
2
3
4
5
6
7
8
11U
81-82/"
108 -
106
-
104
-
^^X^Average
_
102
^
f
^
a w %
^^
100
-
98
Oft
•
•
i
i
i
-6
-5
-4
-3
-2
i
-1
i
T
1
Quarters
•
i
i
i
2
3
4
5
Note: Includes nondurables and services consumption, and government purchases.
AVG includes 1960-61, 1969-70, 1973-75, 1981-82 recessions.
i
structure (discussed below), particularly the end of Regulation Q, make historical comparisons for this sector difficult to interpret.
Chart 8 shows that both federal and state and local government purchases had a depressing effect on the economy that were largely independent of credit conditions and
monetary policy. Here large federal budget deficits and deteriorating revenues at the
state and local level kept these sectors from making their normal contributions to economic activity.
Finally, Chart 9 gives peak and trough comparisons for the trade sector. Peak comparisons show that exports were exceptionally strong, probably a lagged effect of easier
monetary policy and a falling dollar during the late 1980s. Slower export growth in
1991-92 was probably more a reflection of weak economic performance abroad than of
any credit constraints or U.S. monetary policy effects. In fact, monetary policy easing
during 1991 and 1992 probably made U.S. goods even more competitive in international
markets, but slow growth abroad muted this effect.
In sum, a casual look at the aggregate evidence shows extraordinarily weak economic growth, particularly during the so-called economic recovery. The weakness in aggregate demand seems to be fairly evenly distributed among sectors that are sensitive to
credit and monetary policy and those that are not. There is little evidence that monetary
policy was completely ineffective; relative to their normal cyclical patterns, policy-sen-
Chart 5: Real Consumer Spending
Peak = 100
115
Durables
Peak =100
110
104
•
105
102
-
100
100
Nondurables and Services
Average^^
-
98
Average
90
f81-82
<
90-91
95
y
-4
-3
-2
-1
P
1
2
3
4
5
•
6
7
8
-4
•
-3
•
-2
•
-1
P
•
i
•
•
•
•
•
•
1
2
3
4
5
6
7
8
1
2
3
4
5
6
Trough = 100
108
Nondurables and Services
Trough = 100
130
Durables
81-82^
120 •
f
Average
110
<&—/
100
90
-6
-5
-4
-3
-2
•
i
i
i
i
•
-1 T
1
Quarters
2
3
4
5
6
-6
-5
-4
-3
-2
-1
T
Note: AVG includes 1960-61, 1969-70, 1973-75, 1981-82 recessions.
265
Causes and Consequences
Chart 6A: Producers' Durable Equipment Investment
Peak = 100
105
100 -
90 -
85
u
-4
-3
-2
-1
P
1
2
3
4
5
6
7
6
7
Quarters
Note: AVG includes 1960-61, 1969-70,1973-75, 1981-82 recessions.
Chart 6B: Nonfarm Inventory to Final Sales Ratio
266
Peak = 100
104
102 100 -
98 -
-4
-3
-2
-1
P
1
2
3
Quarters
4
5
Note: AVG includes 1960-61, 1969-70,1973-75, 1981-82 recessions.
8
Chart 7A: Nonresidential Construction Investment
Peak = 1 0 0
100
95
90
X
85
80
7K
\
•S81-82^
90-g i*V-
-
1
1
1
1
-4
-3
-2
-1
1
P
1
1
1
2
3
Quarters
1
1
4
5
1
6
1
7
1
8
Note: AVG includes 1960-61, 1969-70,1973-75, 1981-82 recessions.
Chart 7B: Residential Construction Investment
Peak = 100
81-82/*
120 110
W
100
jf^T-
e
90 80 7n
1
1
1
1
-4
-3
-2
-1
P
i
i
i
i
i
i
i
i
1
2
3
4
5
6
7
8
Quarters
Note: AVG includes 1960-61,1969-70,1973-75, 1981-82 recessions.
267
Causes and Consequences
sitive sectors (with the notable exception of large-scale real estate) may have done a bit
better than the rest of the economy. Nonetheless, subpar economic growth in a world of
easing monetary policy and falling interest rates suggests that policy probably did not
affect the real economy as it had in the past. The next section of the paper begins the
discussion of what role the unprecedented slowdown in credit formation may have
played in this process.
B. The Credit Slowdown
The basic facts about the credit slowdown from 1989 to 1992 can be seen in Table 1.
Chart 8: Real Government Purchases
268
Peak = 100
112
Federal
110
81-82 y
108
106
104
^r
102
^ymmmmmm\K^
/
v^C
^_.«*«—**"*'
100
Average
1
^**^
98
90-91
/
96
•
94
•
-4
-3
Peak = 100
108
•
-2
•
I
•
•
•
•
•
•
•
i
-1
State and Local
Average^^*0'*"
106
104
>
^"^
102
^g!^L
90-91 . . ^ ^
m
t^r^^
"8i*-82
100
—
^
98 96
ui
i
i
i
i
-4
-3
-2
-1
P
1
2
3
Quarters
4
5
Note: AVG includes 1960-61, 1969-70, 1973-75,1981-82 recessions
6
7
8
Comparing pre-recession,4 recession, and recovery periods, growth of both real output
and most credit aggregates was slower than during previous cycles. In particular, nonresidential construction and correspondingly, business mortgages were extraordinarily
weak, declining through most of the 1989-92 period, in contrast to the strong growth in
previous episodes. The performance of equipment and inventory spending was less dramatic, but still was reflected in low growth rates for short-term business credit. Household debt formation was also subpar, mirroring consumption growth. Only residential
construction and household mortgages retained their normal cyclical patterns. General-
4
The pre-rcccssion periods roughly coincide with periods of tight credit. See Eckstein and Sinai (1986),
Owens and Schrcft (1992), Romer and Romcr (1990), and Ryding (1990) for different credit crunch dates.
Chart 9: Real Exports and Real Imports
Peak = 100
115
110 -
-4
-3
-2
-1
P
1
2
3
4
5
6
7
8
Imports
115 -
81-82
/*-
110 105 100
"V
95
s
J^
s S | ( # ^^ l / Average
^*90-91
90 y
-3
-1
P
1
2
3
4
5
Quarters
Note: AVG includes 1960-61, 1969-70, 1973-75, 1981-82 recessions
269
Causes and Consequences
Table 1: Growth Rates over Booms, Recessions, and Recoveries
270
Recessions6
Boomsa
Recoveries6
Past
Episodes
1989-90
Past
Episodes
1990-91
Past
Episodes
1991-92
GDP
2.2
0.9
-2.9
-0.9
5.6
1.6
Debt-sensitive sectors
2.3
-1.4
-9.4
-2.1
9.4
2.5
Durables consumption
-1.3
-2.9
-6.8
-2.9
19.0
5.0
P.D.E.
4.5
-0.4
-11.7
-3.0
7.8
1.8
Nonresidential structures
8.6
1.3
-8.3
-4.0
-2.5
-10.4
-10.1
-10.8
-13.9
-7.0
32.8
13.4
Inventories (stock)
3.9
2.0
1.1
-0.7
1.3
-0.6
All other sectors
-0.1
2.3
6.5
1.2
-3.8
-0.9
Private nonfinancial debt
11.8
6.7
10.8
0.6
8.8
2.3
Business debt
13.4
4.1
13.7
0.3
8.2
-0.6
Short-term credit
17.0
5.8
1.0
-1.2
4.7
-0.8
Loans
18.9
4.5
14.1
-0.5
4.3
-4.6
Bank loans
18.2
1.2
14.1
-0.1
1.8
-5.4
Business mortgages
12.5
1.5
12.1
-0.4
11.9
-2.6
Household debt
11.4
9.3
7.6
0.8
9.5
4.6
Mortgages
10.8
11.2
8.0
1.3
9.3
5.8
M1
5.8
4.6
8.3
1.0
7.6
10.9
M2
6.6
5.3
10.3
0.8
13.1
2.8
Fed funds
4.28
-0.92
-5.12
-1.73
-0.72
-2.40
Prime rate
3.61
-0.66
-3.49
-0.81
-1.57
-2.69
Inflation rate
3.57
0.94
-1.31
-0.28
-2.69
-2.42
Long-term govt. bond rate
1.60
0.60
-1.28
-0.69
0.09
-0.71
Real growth rates
Residential structures
Growth rates
Differences
a
Includes four quarters up to NBER peak dates. Past episodes are average of the booms preceding the 1969-70, 1973-75 and 1980 recessions.
b
Past episodes include average of 1969-70, 1973-75 and 1981-82 recessions.
c
- Recoveries are NBER trough to four quarters later. Past episodes are average of recoveries
after the 1969-70, 1973-75 and 1981-82 recessions.
ly speaking, intermediated credit, particularly bank loans, slowed more sharply than
market credit, and the broad money aggregates were correspondingly weak.
Table 1 also points out the most unusual characteristic of the "credit crunch" and the
recession: they were not precipitated by a sharp tightening of monetary policy.5
Instead, the slowdowns (and eventual declines) in both credit and activity appear to
have been more closely related to several credit demand and supply factors: (1) Debt
overhang from the 1980s and the associated strain on nonfinancial balance sheets appears to have caused a fundamental downward shift in credit demand and economic activity, particularly in the consumer sector; (2) Structural changes in several areas of
business investment, particularly in computers and inventories, caused an exogenous
decline in credit demand unrelated to balance sheet positions; (3) Economic activity was
also curtailed by credit supply problems, which were largely the result of the poor financial health and balance sheet problems of many intermediaries.
Although much attention has been paid to the third factor, credit supply constraints
and their effect on monetary policy transmission, several studies have found a substantive role for balance sheet adjustments and demand shifts in the overall credit slowdown. Because credit supply problems are, in practice, difficult to distinguish from
balance sheet restructuring by consumers and nonfinancial firms, the following sections
discuss the three credit factors in more detail and introduce several credit proxies that
attempt to measure them.
1. Balance Sheet Effects
The deleveraging/balance sheet story is a straightforward one. During the 1980s there
was a huge leveraging up by households and businesses. By the late 1980s the increased
leverage, combined with relatively high real interest rates, created historically high debt
burdens that cut sharply into consumers' discretionary spending and firms' cash flow.
For consumers, increased leverage combined with stubbornly high consumer interest
rates resulted in record high debt service burdens by the latter half of the 1980s, leaving
less income for discretionary consumption. Changes in the tax treatment of interest payments combined with the generally overburdened household sector led to deleveraging
late in the decade, a development that put further downward pressure on consumption.
This story emphasizes the income effect of high leverage and is supported by the evidence from the composition of consumer spending in Chart 5 above: sectors that are relatively insensitive to credit, such as nondurables and services consumption, experienced
growth just as anemic (relative to historical norms) as that in more credit-sensitive sectors such as consumer durables. Thus, high debt service helped to reduce spending
growth in a wide variety of goods, not just those purchased with credit.6
Leverage effects on firms created similar problems. Both theoretical (Jensen and
Meckling 1976) and empirical (Fazzari, Hubbard and Peterson 1988) studies have
shown that the investment and production plans of high-leverage firms respond more
strongly to bad demand shocks than do the plans of firms with lower leverage. The macroeconomic implications of these leverage effects for both investment and credit are
5
Although policy was tightened in 1988, the tightening was short-lived, and short-term rates were falling
by late 1988. Declining short-term rates and the absence of the Regulation Q ceilings mean that the credit
slowdown and ensuing recession cannot be attributed to disintermedialion-style credit rationing, a common feature of historical credit crunches.
6
Further anecdotal evidence suggests that credit supply constraints may have been less important because
banks and financial institutions were attempting to expand certain types of consumer lending, particularly
credit cards, during this period.
271
Causes and Consequences
clear: as the business sector became more highly leveraged over the 1980s, investment
and credit demand reacted more strongly to declines in economic activity. As with
households, any attempt by firms to deleverage further depressed investment and credit
demand.
2. Credit Demand Shifts
In addition to debt overhang problems, several credit- (and policy-) sensitive components of aggregate demand underwent secular changes in the 1980s that kept growth in
these components and their related credit demands low relative to historical norms. The
paper by Mosser and Steindel in this volume emphasizes two particular sectors, business
investment in equipment and inventory investment, that underwent structural changes
largely unrelated to credit market developments. As the relative price of computers fell
sharply in 1980s, firms shifted expenditure away from production equipment and toward the relatively cheap computers, thus reducing nominal equipment growth and related credit demands.7 Similarly, improved inventory management techniques,
particularly by manufacturers, helped to keep the inventory cycle and associated credit
demands quite small in the 1990-91 recession. Because these structural changes appear
to predate the slowdown in credit in the late 1980s, it seems likely that they reduced
credit demand and thus contributed to, although by no means completely explain, the
slowdown in short-term business credit.8
3. Credit Supply Effects
272
Several recent studies have examined the credit slowdown, attempting to control for
shifts in credit demand factors and some of the balance sheet effects noted above. In
general, they conclude that credit supply restrictions, particularly from banks, were responsible for a portion of the credit slowdown. For example, Bernanke and Lown
(1991) concluded that although deteriorating balance sheet positions of consumers and
nonfinancial firms were responsible for most of the slowdown in activity and credit, the
supply of bank credit, particularly in certain regions, was an important factor impeding
the economy. In this volume, Lown and Wenninger estimate that bank loans were significantly weaker than traditional reduced form lending-activity relationships would indicate. Cantor and Rodrigues find similar results for nonbank intermediated credit.
Mosser and Steindel show that the long-term relationships between credit flows and
economic activity fell apart in the late 1980s, with credit flows nearly 35 percent too low
given the pace of economic activity.
More microeconomic studies, such as Johnson (1991), Peek and Rosengren (1992),
and Hancock and Wilcox (1992), suggest that bank capital problems, probably related
to the real estate collapse, restricted credit supply in excess of what interest rates and
demand factors would have predicted. In this volume, Hamdani, Rodrigues and Varvatsoulis use survey evidence to show that credit supply restrictions were particularly important for smaller firms, perhaps because of their almost exclusive reliance on
intermediated credit.
Although nearly all of the studies agree that a "capital crunch" story appears to fit the
7
In addition, deflating nominal credit by the GDP deflator (as is common practice) makes real credit growth
look disproportionally weaker than investment demand.
8
In fact, Mosser and Steindel find that fundamental demand factors explain only about 25 percent of the
decline in short-term corporate borrowing from 1989 to 1991.
general facts of the credit slowdown, estimates of the size of the credit supply restrictions vary. Most of the macroeconomic studies mentioned above measure credit supply
restraints using one or more proxies, including interest rate spreads, measures of credit
restrictions for small firms, and regression residuals that measure excessive credit restraint.
In their paper, Lown and Wenninger discuss several proxies for credit supply restrictions from the banking sector.9 For example, spreads between bank borrowing rates and
market rates appear to have been kept high in an attempt to discourage borrowing. One
signal of the reluctance of banks to lend was a 250+ basis point gap between the prime
and federal funds rates that persisted over 1991 and 1992. In contrast, during previous
recessions and crunches the spread between the prime and fed funds spiked to high levels when inflation was high and monetary policy was tightened, but returned to more
normal levels relatively quickly.
Lown and Wenninger also show that reduced form relationships between different
components of bank lending, economic activity, and borrowing rates overpredicted
lending during the crunch. Although their equations are not, strictly speaking, structural
loan demand equations, they largely capture credit demand effects. Below, residuals
from their estimates are used as proxies for bank credit supply constraints.10
Another set of credit "supply" proxies come from the paper by Mosser and Steindel
in this volume. They show that long-term relationships between a number of credit aggregates (both bank and nonbank credit), interest rates, and economic activity consistently underpredicted actual credit flows in the mid 1980s and overpredicted flows in the
early 1990s. While their regressions are not credit demand relationships in the formal
sense, in practice they behave suspiciously like stable, long-term demand relationships.
Thus their large residuals reflect either criedit supply shifts or some nonlinear credit demand changes perhaps related to deleveraging.11
Finally, Hamdani, Rodrigues and Varvatsoulis suggest proxies for credit supply constraints that focus on small firms. Results from Gertler and Gilchrist (1992) and Oliner
and Rudebusch (1992) suggest that because small firms have little or no direct access to
capital markets, their activities are harder hit by credit restrictions at financial intermediaries than the activities of large firms. Using survey results from the National Federation of Independent Business, Hamdani, Rodrigues and Varvatsoulis find that even
after adjusting for normal cyclical movements in output and interest rates, many small
firms had significant difficulty in obtaining credit in the late 1980s. Their proxy essentially measures the amount of nonprice credit rationing faced by small firms.
9
Lown and Wenninger offer extensive evidence in favor of this bank capital/credit supply story. They show
that virtually all categories of bank lending (excepting home mortgages) were weaker than in past downturns. In addition, bank surveys, changes in lending requirements (for example, higher collalcrali/ation)
and substantially weaker lending by the lowest capital banks all suggest that constraints on bank credit
supply were important. Finally, their econometric estimates of relationships between bank lending, interest rates, and economic activity suggest that the weakness in bank lending (particularly consumer loans
and nonresidential mortgages) in the 1990-91 recession was substantially different from that in previous
economic downturns.
10
Of course, Lown and Wenninger's residuals, like those from Mosser and Steindel below, may also reflect
some nonlinear credit demand effects. For example, higher leverage may have changed the historical relationship between spending and credit demand. In other words, at high debt levels, cuts in spending may
cause proportionally larger declines in credit demand as firms and households deleverage and cut spending
simultaneously.
11
More precisely, their residuals measure those balance sheets changes that are uncorrelated with economic
activity.
273
Causes and Consequences
The studies examined here provide ample evidence that shifts in both credit demand
and supply played an important role in the credit slowdown. The next question is what
effect the credit slowdown, and credit supply restrictions in particular, may have had on
aggregate activity. As a first step to answering that question, the next section of the paper measures the statistical relevance of credit variables, including the supply proxies
discussed above, in explaining systematic movements in aggregate economic activity.
Further econometric evidence is discussed in section III below.
C. Granger-Causality Tests of Credit and Aggregate Demand
274
This section of the paper looks at the time series relationship between credit and economic activity during the 1980s. Specifically, it reports Granger-causality tests of
whether overall credit aggregates or credit supply proxies are statistically significant
predictors of future economic activity. These statistical tests are based on data from
1980 to 1992, and thus are useful indicators of systematic links between credit, credit
supply proxies and future activity over that period. To the extent that the 1989-92 credit
slowdown was a unique episode, however, these tests may inadequately measure the impact of credit on activity. Thus these tests should be viewed as only an introduction to
the linkages between credit and activity.
Following the previous literature on the relationship between credit and economic
activity,12 Tables 2 and 3 present Granger causality tests statistics from regressions of
economic activity (real GDP and debt-sensitive real GDP) on lagged credit variables
and lagged activity. The F-statistics test whether all the coefficients on lagged credit
variables are significantly different from zero. Figures in parentheses are significance
levels of the F-tests (percentiles of the F-distribution). Table 2 reports tests for regressions with eight lags of activity and credit; regressions in Table 3 also include eight lags
of the federal funds rate.
Several results stand out from the tables. Aggregate credit (and money) variables
were generally poor predictors of activity, although M2 was the best of a bad lot.13 In
general, the proxy variables did a better job of explaining economic activity in the 1980s
than credit or money aggregates. The bivariate results show that the prime rate-federal
funds rate spread, small firm borrowing constraints, and residuals from Mosser and
Steindel debt regressions all Granger-cause activity. In light of the emphasis on bank
behavior and the real estate sector in the credit slump, the only surprising result is that
residuals from business mortgage regressions and from Lown and Wenninger's bank residuals were generally insignificant in explaining activity. In addition, all money, credit,
and proxy measures were less important in explaining activity if the federal funds rate
was also used as an explanatory variable.14 Even so, the credit proxies in the three variable regressions were substantially more important in explaining economic activity than
were the corresponding credit aggregates.
Overall, this formal econometric evidence using Granger causality tests is somewhat
12
These tests arc closely related to those used by Bcrnanke and Blinder (1992), Friedman and Kuttncr
(1989), Romcr and Romer (1990) and Kashyap and Stein (1992), to determine if monetary policy affected
real activity more through a "money" channel (interest rates) or through a "credit" channel. See
footnote 2.
13
All money and credit variables are worse predictors of activity in the 1980s than during earlier periods,
probably as a result of the massive structural changes in financial market and intermediaries during the
decade.
14 This finding has been standard since Sims (1972).
Table 2: Granger Causality Tests: Credit, Credit Proxy, and Real Activity8
Differences, 1980 to 1992
Real GDP
Debt-sensitive GDP
Nonmortgage debt
1.05(0.39)
1.47(0.23)
Short-term credit
1.65(0.18)
2.47 (0.06)*
Loans
0.94 (0.45)
1.04(0.49)
Consumer credit
1.45.(0.24)
2.64 (0.05)**
M1
1.11 (0.36)
1.41 (0.25)
M2a
2.02(0.10)*
2.21 (0.09)*
Nonmortgage Corp. Debt
2.96 (0.03)**
2.30 (0.08)*
Short-term
2.68 (0.04)**
5.35 (0.00)***
Loans
1.13(0.35)
0.81 (0.52)
Business mortgages
0.99 (0.42)
1.07(0.38)
Consumer credit
1.28(0.29)
1.52(0.21)
Household mortgages
1.46(0.24)
1.93(0.12)
Consumer loans
0.23 (0.92)
0.49 (0.74)
C&l loans
1.51 (0.22)
2.82 (0.04)**
Business mortgages
1.20(0.33)
0.32 (0.86)
Bivariate
Mosser/Steindel residuals'3
Lown/Wenninger residuals0
Hamdani/Rodrigues/Varvatsoulis residualsd
All firms
Regular borrowers
Bank spreade
2.10(0.10)*
1.85(0.14)
2.69 (0.04)**
2.54 (0.05)**
2.34 (0.07)*
3.64(0.01)***
a
- Debt-sensitive real GDP is consumer spending on durables, business fixed investment and
inventory investment. Statistics reported in the table test the null hypothesis that all lags of
credit proxies have zero coefficients in explaining real GDP. Numbers in parenthesis are percentiles of appropriate F-distributions.
b
- Residuals from Mosser and Steindel regressions of the long-term relationships between debt
and economic activity.
c
- Residuals from Lown and Wenninger reduced-form regressions relating bank loans to economic activity and interest rates.
d
- Residuals from Hamdani, Rodigues and Varatsoulis reduced-form regressions relating small
business borrowing constraints (as measured by survey data from the NFIB) to economic
activity and interest rates.
e
Prime rate minus federal funds rate.
* Statistically significant at 10 percent level.
** Statistically significant at 5 percent level.
*** Statistically significant at 1 percent level
275
Causes and Consequences
Table 3: Granger Causality Tests: Creidt, Credit Proxies, and Real Activity3
Differences (1980-92)
276
Real GDP
Debt-sensitive GDP
Nonmortgage debt
0.78 (0.54)
1.00(0.42)
Short-term credit
0.61 (0.66)
0.84(0.51)
Loans
0.48 (0.75)
0.58 (0.68)
Consumer credit
1.11 (0.36)
2.36 (0.07)*
M1
1.16(0.35)
0.86 (0.49)
M2a
0.31 (0.87)
0.45 (0.77)
2.96 (0.03)**
2.43 (0.07)*
Short-term
0.51 (0.73)
1.76(0.16)
Loans
0.77 (0.55)
0.49 (0.74)
Business mortgages
0.17(0.95)
1.08(0.38)
Consumer credit
0.45 (0.77)
0.34 (0.85)
Household mortgages
0.99 (0.43)
1.89(0.14)
Consumer loans
0.54(0.71)
0.03 (0.99)
C&l loans
0.49 (0.75)
1.62(0.19)
Business mortgages
1.25(0.31)
0.81 (0.53)
Trivariate with federal funds rate
Mosser/Steindel residualsb
Nonmortgage Corp. Debt
Lown/Wenninger residuals0
Hamdani/Rodrigues/Varvatsoulisresidualsd
All firms
1.92(0.13)
1.56(0.20)
Regular borrowers
1.86(0.14)
1.70(0.17)
1.18(0.33)
1.50(0.22)
Bank spreade
a
Debt-sensitive real GDP is consumer spending on durables, business fixed investment and
inventory investment. Statistics reported in the table test the null hypothesis that all lags of
credit proxies have zero coefficients in explaining real GDP. Numbers in parenthesis are percentiles of appropriate F-distributions.
b
Residuals from Mosser and Steindel regressions of the long-term relationship between debt
and economic activity.
c
Residuals from Lown and Wenninger reduced-form regressions relating bank loans to economic activity and interest rates.
d
Residuals from Hamdani, Rodigues and Van/atsoulis reduced from regressions relating small
business borrowing constraints (as measured by survey data from the NFIB) to economic
activity and interest rates.
Prime rate minus federal funds rate.
* Statistically significant at 10 percent level.
** Statistically significant at 5 percent level.
e
disappointing. Although credit supply proxies are more important in explaining activity
than credit aggregates, few of the credit variables are statistically significant, particularly when the federal funds rate is included. Taken at face value, the Granger-causality
tests suggest that the quantity and composition of credit had little or no independent effect on output during the 1980s. In contrast, the informal evidence on aggregate demand
presented earlier suggested a very close relationship between credit and activity.
One reason for the ambiguous econometric results may be that aggregate regressions
and statistical tests mask a number of important sectoral differences in the interaction
between credit formation and economic activity. Both institutional details and the large
differences across the aggregate demand components in Charts 5-9 suggest that relationships between activity, credit and policy for real estate, households, and businesses
should be examined individually. Section III below does this, using the credit proxies
to help explain the weakness in different demand components from 1989 to 1992.
Before turning to sectoral evidence, however, a second problem associated with the
statistical tests above needs to be discussed. Both the simple recession comparisons and
the Granger-causality tests abstract from a number of important structural differences
between the 1990-91 business cycle and its predecessors. In addition to the extraordinary conditions in credit markets, these differences include the stance of monetary policy, financial innovation and deregulation that have changed the transmission of
monetary policy, fiscal policy stance, and other "exogenous" shifts in aggregate demand. Section III below looks at the role of these factors (in conjunction with the credit
slowdown) in curtailing economic growth, with particular emphasis on monetary policy
transmission.
Factors Influencing Monetary Policy Effectiveness
This section of the paper looks at the linkages between monetary policy and the real
economy, giving particular attention to those factors, including the credit crunch, that
may have disrupted or changed these linkages. Section A compares monetary policy
stance during the 1990-91 cycle and earlier cycles. Section B reviews how financial
market problems in the early 1990s, particularly the credit slowdown, may have made
monetary policy less effective. Section C summarizes the long-term changes in monetary policy transmission mechanisms that may have altered the policy-credit-output
linkages. Section D discusses other exogenous factors that may have offset monetary
policy easing.
A. Monetary Policy Stance
Before discussing the interaction of monetary policy, activity and the credit slowdown
in the early 1990s, some comparison of monetary policy stance with that in earlier cycles is necessary. Although the 1990-91 recession began with monetary policy easing
rather than tightening, sluggish growth in both credit and the real economy might have
occurred because monetary policy did not ease enough to stimulate activity. One reason
for confusion about the "tightness" of policy is that the two traditional measures of monetary policy stance, M2 and the interest rate on federal funds, gave conflicting signals in
the early 1990s.
M2 growth suggested a tight policy stance (Chart 10A). The slow growth in M2,
however, was due to sharp declines in small certificates of deposit, and appears to have
been a conscious choice on the part of financial institutions to reduce managed liabilities
in the face of capital problems. Thus, M2 behavior reflected the capital crunch at de-
277
Causes and Consequences
Chart 10A: Real Money Growth
Percent
20
15
1 1
i
I
10
5
0
-5
-10
-15
y
-I
1960 62
^
s
I
I
RealMI ^
Ii
AM
liar I
JA I
ill
f
iff ^
V
I
1 1.
11,
1
•
i
i
i\\
i
64 66
68
70
72
74 76 78 80 82 84 86 88 90 92
Chart 10B: Real Federal Funds Rate
278
1968 70
72
76
78
80
82
84
86
88
90
92
Note: The real Federal funds rate is calculated as the nominal rate minus the previous
twelve months inflation in the core CPI.
pository institutions but was not a clear signal of monetary policy.
In contrast, the real, or inflation-adjusted federal funds rate (Chart 10B) fell to near
zero by mid-1992, suggesting policy was looser than it had been in the 1981-82 recession, but a bit tighter than at the end of earlier recessions. Nonetheless, the real funds
rate did rise more than 150 basis points in 1991 when nominal rates were flat and inflation slowed sharply. This "induced" tightness, which resulted from the Fed's policy of
very gradual easing, suggests that monetary policy stimulus only slowly worked its way
through the economy.
B. "Blockages" in Monetary Policy Transmission
Although monetary policy stance from 1989 to 1992 was somewhat different from past
cycles, policy stance alone cannot completely explain the extraordinarily slow growth
in real activity and intermediated credit. Indeed both the credit supply and demand stories outlined above suggest that during this period the transmission of policy changes
via financial markets and institutions was weaker than in past episodes.
Some of the channels through which policy should affect the economy, but which
may have been blocked because of financial system fragility, are shown in Figure 1.
Blockage (1) suggests that the capital crunch and banking crisis did not allow reductions
in short-term market interest rates to be transmitted to bank lending rates or credit
Figure 1: Channels of Monetary Policy
"(3,
Bank
Lending
Money
\
/
Increase in
Reserves
Lower Short
Rates
|
1
X
\
Dollar
Depreciation
Business eind
Househo
a
Spendin
3
Lower Long
Rates Higher /
Asset Prices
Trade
Possible "Blockages"
(1) Large writeoffs, new capital requirements, poor asset quality (balance sheet problems).
(2) Uncertainty, inflation concerns, long-term budget deficit.
(3) Deleveraging and balance sheet problems of firms and households.
279
Causes and Consequences
growth. Evidence of this shift seems quite overwhelming: credit growth, particularly
at depository institutions, was well below economic fundamentals and bank lending
rates relatively high.15
Blockage (2) suggests that the channels to long-term bonds and other assets were
blocked by inflation fears or by a high level of investor uncertainty—perhaps related to
credit problems or to concerns about long-term federal deficits. Certainly, long-term
rates fell substantially less than short-term rates, making the term structure quite steep,
but the term structure is usually steep when the economy is weak and monetary policy
is loose (Chart 11). How important the credit slowdown and banking crisis were in the
steep yield curve is unclear however, since the medium-term spread between three
months and three years was within historical norms, but the long-term yield curve (three
to ten years) was extraordinarily steep.
Blockage (3) suggests that deleveraging and debt restructuring by households and
firms depressed spending and credit demand, making them less responsive to easing
monetary policy. For example, lower interest rates due to easy monetary policy may
have induced firms and consumers to deleverage more quickly, while keeping spending
growth modest.16 Although easing monetary policy may have "worked" by averting a
deeper and more prolonged recession, this pattern is in sharp contrast to historical experience, when declining interest rates encouraged increases in economic activity. Thus
the normal pattern of easing monetary policy followed by a strong rebound in economic
activity appears to have been interrupted by shifts in both credit supply and credit demand.
C. Long-Term Changes in Monetary Policy Transmission
280
The short-run impediments to policy discussed above are only part of the reason that the
relationship between monetary policy, credit, and economic activity in the early 1990s
was so different from historical experience. Since the late 1970s, both financial innovation and deregulation have not only changed the behavior of financial institutions and
markets, but they have also fundamentally altered the relationship between financial
markets and economic activity. Even in the absence of the credit "crunch," these longterm changes in the channels of monetary policy transmission would have altered both
the size and speed with which policy affected various sectors of the economy.
Tables 4 and 5 summarize by sector and by study some of the major findings on longterm changes in monetary policy transmission. Overall, these studies conclude that
monetary policy today is transmitted less through Regulation Q-type quantity rationing
of credit and more through interest rates and their effect on balance sheets and the ex15
Sec the paper by Lown and Wenninger and Table 1. There is a small caveat to this story. Although bank
rates remained high relative to market rates in 1991 and 1992, in fact bank lending rates (with the notable
exception of consumer credit rates) did decline steadily from late 1990 to the middle of 1992. It appears
that the way in which monetary policy was "working" was to allow banks sufficient liquidity to improve
their balance sheet positions in a relatively short period of time. While this outcome of monetary policy
did not lead directly to strong growth in real activity, tighter policy might have led to large-scale bank failures and financial distress, and certainly to a deep and prolonged recession.
16
Tighter monetary policy and higher rates would have forced further cuts in discretionary purchases as
agents attempted to lower their debt burdens. This argues that easy monetary policy prevented a short-run
collapse in spending and sped up deleveraging at the same time. This process seemed to be particularly
clear in the household sector. There, lower interest rates led to a record volume of mortgage refinancing,
freeing up funds for both discretionary spending and for further retirement of debt. In addition, lower
rates stimulated housing investment and related consumer durables spending.
Chart 11: Treasury Spreads
Percent
10 Year Bond - 3 Month T-bill
92
I1
3 Year Note - 3 Month T-bill
4
3
i"
2
1
0
-1
-2
1967
72
82
77
87
92
Percent
x"
10 Year Bond - 3 Year Note
^
$^
\ \ \
1967
VVVv
72
^
1
^
1 te
x
1 I
-1 -2
^
77
^\s x
11
82
1 "
i
87
^
i
92
Note: Shaded areas represent recessions.
281
Causes and Consequences
change rate.17 Roughly speaking, the empirical evidence suggests that a smaller response of housing to monetary policy has been offset (by some measures, more than
offset) by a stronger exchange rate and net export response to policy. Although the studies in Tables 4 and 5 differ in their estimates of the overall size of the change in monetary
policy effectiveness, they agree that the speed with which monetary policy affects the
real economy is slower, and the impact of policy in any particular month or quarter is
probably more unpredictable.
From the perspective of the 1990-91 cycle, longer and more uncertain policy lags
make it very difficult to distinguish between short-run blockages to policy caused by
credit supply and demand shifts and the longer term changes in transmission mechanisms. For example, monetary easing in 1991 and 1992 may have weakened the dollar,
eventually giving a boost to U.S. exports; however, other long-term structural changes
and the recent credit crunch may have conspired to make monetary policy less effective
in stimulating domestic demand. Furthermore, if policy is now transmitted via interest
rates and their impact on balance sheets, then balance sheet restructuring by consumers
and firms could short-circuit monetary policy easing. In addition, credit supply shifts
caused by bank capital problems could still keep easier monetary policy from affecting
those sectors most dependent on intermediated credit despite wider capital market access for many firms and households. Large-scale real estate appears to be one such sector.
D. Other Factors Offsetting Monetary Policy
While changes in transmission mechanisms and short-term credit problems may have
muted monetary policy effectiveness in the latest cycle, economic growth was also re17
This conclusion does not mean that credit rationing by intermediaries would not affect the transmission of
monetary policy, only that Regulation Q-typc interest rate ceilings, which induced systematic credit
rationing under very tight monetary policy, arc no longer important.
Table 4: Summary of Empirical Studies
Changes in the Transmission of Monetary Policy
282
Final Demand Components
Innovation/Deregulation
Monetary Policy Influence
Housing
Remove Reg. Q
Mortgage securitization
Variable rate mortgages
Business investment
Increased leverage
Greater access to commercial Likely larger, but few studies
paper and loan commitments using aggregate data
Inventory investment
Greater access to commercial
Possibly larger, but few studies
paper and loan commitments
Consumption
Increased leverage
Greater credit access/info.
Wider menu of saving instruments
Trade
Less for large shocks; maybe
more for small changes
No discernible change
Reduced capital controls
Swaps and futures
"Internationalization" of capital Larger impact of interest rate
markets
differentials on the dollar
strained by "headwinds"—negative factors largely unrelated to monetary policy stance.
These headwinds include relatively tight fiscal policy or rather lack of any significant
fiscal stimulus, and low confidence or "animal spirits", particularly by households18
Chart 12A makes clear that discretionary fiscal policy, measured by the full employment budget deficit, eased much less than in previous recessions. There was about a
quarter of the discretionary fiscal pump priming done in previous recoveries and only
one-sixth of the discretionary stimulus done in 1981-82. This lack of fiscal stimulus is
unique: the U.S. did not have a single sustained economic recovery in the preceding
18
In addition, the two previous recessionary periods, 1973-75 and 1980-82, were preceded by extremely
large (negative) supply shocks (mainly large energy price increases). The subsequent easing of these supply shocks was a factor in the economic recoveries that followed. In the latest episode, the energy price
shock was small and brief, playing a modest role in recession and recovery.
Table 5: Empirical Studies of Monetary Policy Transmission
Sector
Study
Major Conclusions
Akhtar & Harris: FRBNY
Aggregate
Interest sensitivity of private spending
has risen, credit rationing less and
exchange rate effects are greater
Bernanke & Campbell:
Brookings
BFI
Serious risk to real economy (investment) from leveraging up in 1980s
Bosworth: Brookings
Housing ;consumption;
trade
Housing less sensitive; others perhaps
more, but policy lags are longer
Cantor: FRBNY
BFI
Investment of highly leveraged firms
more closely tied to sales and cash flow
Fazzari et. al.: Brookings
BFI
Financial structure of firms very important in short-run investment decisions
Friedman: FRBKC
Aggregate
Less housing sensitivity; greater BFI
sensitivity; little change in consumption
Hirtle: FRBNY
Aggregate
Change in relationship between GNP
and interest rates that can be related to
loan commitments
Kahn: FRBKC
Aggregate
GNP slightly less sensitive to Fed funds
rate; longer and uncertain policy lag
Aggregate
MPS model reestimation confirms less
sensitive housing, more sensitive trade
sector, but little overall change in policy
sensitivity
Mosser: FRBNY
Aggregate
Large macroeconometric models are
more sensitive to policy than in the
1970s; policy lags are longer; housing
slightly less sensitive, trade much more
Ryding: FRBNY
Housing
Less sensitivity to policy due to decline
in disintermediation
Throop: FRBSF
Housing
Less sensitive to policy, but size of the
effects are small
Mauskopf: FRB
283
Causes and Consequences
Chart 12A: Change in the Full Employment Budget Deficit
Business Cycle Peak to Eight Quarters Later
Percent of Potential GDP
2.5
1970
1975
1982
Source: Federal Reserve Board.
Chart 12B: Consumer Sentiment
Michigan Survey
284
Peak = 100
120
110 -
100 -
80
LL
-4
-3
-2
-1
Note: AVG includes 1969-70, 1973-75,and 1981-82 recessions.
1990
thirty years without significant fiscal stimulus. Indeed, a fiscal stimulus package comparable to that implemented in the early 1980s might have added a full percentage point
to GDP growth in 1991 and 1992.19
Economic activity also appears to have been hurt by low consumer sentiment. Chart
I2B shows that consumers were very pessimistic in late 1991 and 1992, more so than in
previous economic recoveries. The low confidence readings may have been partly related to the credit slowdown, however, pessimistic confidence readings also stemmed
from long-run changes in consumer prospects. In particular, corporate restructuring and
downsizing have made future permanent income (and benefits) more uncertain, and a
decade of relatively slow income growth and increasing income inequality (with no end
in sight) probably contributed to the gloomy outlook.
Low confidence and the lack of any fiscal stimulus make the slow-growth recovery
discussed in section I more understandable, regardless of the role of the credit slowdown
and any long-term changes in monetary policy effectiveness. Because monetary policy,
for the first time, was "going it alone" in trying to stimulate the economy, the simple historical comparisons with previous cycles presented above may be misleading indicators
of whether the economy was "on track" in the late 1980s and early 1990s. To evaluate
monetary policy efficacy and the role of credit in the downturn more rigorously, the next
section presents evidence from several econometric models of aggregate activity.
Econometric Evidence on Monetary Policy Effectiveness and the Role of Credit Restraint
One way to pull together many of the factors affecting economic growth is to ask whether final demand equations, which relate spending to economic fundamentals such as income, wealth, interest rates and relative prices, adequately captured the weakness in the
real economy from 1989 to 1992. On the one hand, if the slow economy was largely
due to excessively high long-term interest rates or to relatively tight fiscal policy, or if
balance sheet effects on aggregate demand are reflected largely through interest rates,
these equations should predict economic activity fairly well. On the other hand, if factors left out of the models—for example, credit supply problems or an exogenous shift
in demand for both goods and credit (such as a desire to deleverage)—were responsible
for the weakness, then these equations will perform badly and consistently overpredict
real activity.
This section looks at the performance of a number of different final demand models
during the credit slowdown. In general, both reduced-form estimates and structural
equations from large econometric models overpredicted real spending from 1989 to
1992. Adding the credit supply proxies from section I helps to explain some, but not all,
of the overpredictions in policy-sensitive sectors. In addition, the overpredictions are
not restricted to those sectors most sensitive to policy.
Chart 13 shows that aggregate reduced-form equations generally overpredicted output from 1989 to 1992.20 While policy-sensitive sectors underperformed relative to
model predictions, so did the sectors less directly sensitive to monetary policy. These
models suggest a general malaise in aggregate demand, policy-sensitive or not, which
19
20
This calculation assumes a fiscal multiplier of 1.
Separate prediction models for policy-sensitive and insensitive sectors of real GDP were estimated by
regressing each on four lags of the long-term government bond rate, inflation, monetary policy (federal
funds rate and M2) and a measure of discretionary fiscal policy (the full-employment budget deficit) from
1967 to 1992-11. Prediction errors arc static: actual expenditure less that predicted by the model equation,
using actual historical values for all right hand side variables.
285
Causes and Consequences
was not well described by economic fundamentals.
The general malaise in demand is particularly clear in the consumer sector. Charts
14 and 15 show prediction errors from equations for consumer spending from the DRI
and MPS/Federal Reserve Board models.21 According to the Board model, actual
spending on durables and on nondurables and services was very weak in 1991 and 1992,
well below what income, wealth, and interest rates would indicate. In contrast, DRI's
21
Both DRI and the Board use income, interest rates and relative prices to predict durables spending. In
addition, the Board model uses the slock of durables and the unemployment rale. For nondurables and
services, the Board equation emphasizes household wealth (including the stock market) and permanent
income. DRI uses recent income changes and relative prices.
Chart 13: Reduced-Form Prediction Errors
286
$1987, Billions of Dollars
60
Policy Sensitive Components
.
^
40
20 -
11 '1
vi
0
'I
••\\i
-20
-40
-60 -80
-100
' W
iN s ''
1978
80
82
84
86
88
90
92
$1987, Billions of Dollars
60
Policy Insensitive Components
40 20
i
0
-20
-40
-
I
-60 1-80
1978
80
11
82
I84
Note: Shaded areas represent recessions.
86
88
90
92
equations show consumer spending was about on track because DRI uses consumer sentiment to forecast consumption, while the Board does not. Indirectly, these charts imply
that consumers were more pessimistic than their income, interest rates and wealth levels
warranted.
Chart 16 shows that general reduced-form equations for consumer durables and for
nondurables and services also overpredicted.22 Notably, prediction errors for nondurables and services were, by historical standards, even larger than those for durables. The
22
Both were predicted using disposable income, long-term interest rates, relative prices, inflation, fiscal policy, M2 and the federal funds rate. Four lags of all variables were included in the regression, which was
estimated from 1967 to 1992-11.
Chart 14: Prediction Errors — Durables Consumer Spending
$1987, Billions of Dollars
40
20
-20
-40
-60
1979 80 81
82
$1987, Billions of Dollars
30
83
84
85
86
87
88
89
90
91
92
83
84
85
86
87
88
89
90
91
92
20
10
-10
-20
1979 80
81
82
Note: Prediction errors are actual spending minus model prediction.
287
Causes and Consequences
widespread weakness in consumption suggests that some exogenous factor such as high
debt service burdens or depressed prospects for future income growth caused consumers
to reduce expenditure across the board.
Econometric estimates for the investment sector generally support the informal evidence from Section I. For example, both a reduced-form model and DRI's flexible accelerator equation for nonresidential construction spending overestimated actual
spending substantially (Chart 17). Standard housing equations, both structural and reduced form, also overpredicted slightly (Chart 18). Housing, which was one of the few
areas that appeared to be having a normal recovery in 1991 and 1992, was actually
weaker than what economic fundamentals (demographics, interest rates, income and
Chart 15: Prediction Errors — Nondurables and Services Consumer Spending
288
$1987, Billions of Dollars
i
MPS Model
20 -
a
-20 -
^^
/V
\ s
/
VA
v\
r
-40
•
1979 80
81
82
83
84
85
86
87
88
89
90
91
92
83
84
85
86
87
88
89
90
91
92
$1987, Billions of Dollars
30
1979 80
81
82
Note: Prediction errors are actual spending minus model prediction.
housing prices) would have indicated after mid-1990.
The evidence on producers' durable equipment is mixed. DRI's structural estimates
overpredicted actual levels (Chart 19, top panel), despite the relatively strong performance of this sector in the early 1990s. However, the reduced-form prediction errors
(Chart 19, bottom panel) suggest that equipment spending in 1991 -92 was slightly stronger than what the fundamentals predicted.
Only for exports and imports were traditional model predictions (from DRI and
FRBNY in Chart 20) either too low or on track. Taken together, the relative strength of
the trade sector and the relative accuracy of trade equations are noteworthy because
long-term changes in policy transmission appear to explain the strong trade perfor-
Chart 16: Prediction Errors — Reduced-Form Equations for Consumer Spending
$1987, Billions of Dollars
40
82
1978
84
86
88
90
84
86
88
90
92
$1987, Billions of Dollars
20
Nondurables and Services
1978
80
82
Note: Shaded areas represent recessions,
289
Causes and Consequences
mance, while domestic credit constraints appear to have had little significant effect in
this sector.
To evaluate the role of the credit slowdown in the unexplained weakness in activity,
reduced-form equations for several policy-sensitive sectors were reestimated, this time
including the credit supply proxies discussed in section I above. Comparisons of equations for consumer durables and several investment components, both with and without
the Lown and Wenninger bank loan residuals, are shown in Chart 21. While the business mortgage residuals do improve the performance of the nonresidential structures
equation (particularly in 1990), other Lown and Wenninger residuals make very little
difference in the prediction errors. Both anecdotal and microeconomic studies are consistent with these results: real estate lending (and activity) were most affected by chang-
Chart 17: Prediction Errors — Nonresidential Structures
290
$1987, Billions of Dollars
30
-30
1978
80
82
$1987, Billions of Dollars
1978
80
82
84
Note: Shaded areas represent recessions,
88
90
92
es in credit supply, while evidence of credit supply restrictions for home mortgages and
consumer lending is harder to find. Furthermore, bank credit constraints may have been
important for large-scale real estate spending because borrowers in this sector had few
alternative sources of credit.
Similarly, business mortgage residuals from Mosser and Steindel can explain some of
the overprediction in nonresidential construction (Chart 22). However, the generally positive household mortgage residuals make the overprediction of residential structures even
worse. Like the bank loan residuals, consumer and short-term business credit residuals
make very little difference in predicting consumer durables and equipment investment.
In contrast, the spread between the consumer loan rate and the federal funds rate
Chart 18: Prediction Errors — Housing
Million Units
0.3
DRI Model - Housing Stalls
1978
82
84
82
84
86
88
90
92
$1987, Billions of Dollars
20
1978
80
88
90
92
Note: Shaded areas represent recessions.
291
Causes and Consequences
more than explains the weak consumer durables behavior from 1989 to 1992 (Chart 23).
This finding is clearly the result of ever-widening and record high spreads between consumer interest rates and market rates. However, the changing risk composition and tax
treatment of consumer lending (particularly credit cards) during the second half of the
1980s makes the widening of this spread difficult to interpret as a pure shift in credit supply. For example, the normal cyclical behavior of the spread between credit card rates
and the federal funds rate was to increase from about 5 percent at the peak to about 10
percent at the trough. In the latest recession, the spread started at 10 percent and rose
steadily to over 14 percent by the end of 1992. Chart 23 also shows that the prime rate
spread over fed funds had no additional predictive power for equipment investment.
Chart 19: Prediction Errors — Producers' Durable Equipment
$1987, Billions of Dollars
30
20 10 -
-10 -20 -30 LL
1978
80
$1987, Billions of Dollars
30
82
86
88
90
92
Reduced Form Model
1978
80
82
84
Note: Shaded areas represent recessions
292
88
90
92
Residuals measuring small firm borrowing constraints from Hamdani, Rodrigues
and Varvatsoulis have the largest impact on predictions of business investment components (Chart 23). When these small-firm effects were included in reduced-form equations, they generally lowered predictions of both nonresidential construction and
producers' durable equipment. In fact, the equations for nonresidential construction
switched from overprediction to underprediction when the small-firm credit supply residuals were included. These may be more powerful predictors of business spending
during this period because they were constructed to measure pure (nonprice) credit rationing for small firms (as opposed to shifts in the credit supply curve).
Table 6 summarizes the usefulness of different credit supply proxies in predicting final demand components. Generally speaking, credit supply proxies appear to explain
some of the weakness in investment, particularly in nonresidential construction, where
the largest credit imbalances occurred and where the fewest substitution possibilities exist. However, proxies were less helpful in explaining widespread weakness in other sectors, notably consumption and housing, where the supply proxies were either
unimportant or increased the prediction error (sometimes with the opposite sign).
Unfortunately, the unexplained weakness in the household sector remains a puzzle.
If noncredit factors such as an increase in desired saving or lower expected future income growth are responsible, then slower growth of spending relative to fundamentals
and policy may be permanent, and the final demand relationships estimated above (in-
Chart 20: Prediction Errors — Exports and Imports
$1987, Billions of Dollars
20
DRI Model - Nonaqricultural Exports ; \ \
-20
1987
88
$1987. Billions of Dollars
30
DRI Model - Non-petroleum Imports
91
89
92
$1987, Billions of Dollars
1987
88
89
90
91
92
89
90
91
92
$1987, Billions of Dollars
ID
FRBNY Model
Nonagricultural
10 . Exports
I 1
5
0
-5
1
A
J v \
in •
Kyl l/\A
M #/ y\
if
m
1987
88
89
90
•
91
92
1987
293
Causes and Consequences
eluding links from monetary policy to the real economy) may have permanently
changed. Nonetheless, the timing of the weakness in spending seems too coincidental
for credit factors, particularly balance sheet restructuring, not to have played a role. If
the overpredictions of consumption and housing are the result of "missing" balance
sheet effects, then the shift in aggregate demand may be either permanent or temporary.
For example, if monetary policy now influences activity largely through its effect on
household balance sheets, and if these new balance sheet effects are a result of structural
changes during the 1980s, it seems likely that the relationship between spending, monetary policy, and credit has been changed permanently. Alternatively, if balance sheet
restructuring is simply a onetime adjustment to the debt "bubble" of the 1980s, the shifts
in spending may be temporary. Unfortunately, it is simply too soon to tell which hypothesis is true.
IV. Conclusions
This paper has looked at the relationship between credit, economic activity and monetary policy over the 1989-92 period. Its main conclusion is that the weakness in aggregate demand was more widespread than the weakness in credit. Furthermore, the
Chart 21: Reduced-Form Prediction Errors Including Lown and Wenninger Loan Residual
Actual Less Predicted
$1987, Billions of Dollars
Consumer
Durables
_
A
Mm
20
rv
$1987, Billions of Dollars
25
Producers' Durable Equipment
AJU
With Consumer
Loan Residuals
L
-20
iff
-40 -
1
1988
89
90
Original
With C&l"
Residuals
\VV\VI
91
92
$1987, Billions of Dollars
10
5
1988
89
91
92
$1987, Billions of Dollars
10
Nonresidential Structures
Residential Structures
With Business
(Mortgage Residuals
0
-5
-10
-15
-5
-20
-25
With Household and
Business Mortgage
Residuals
-30
-35
1988
89
90
Note: Shaded areas represent recession.
294
-10
92
-15
Original
1988
89
90
91
92
evidence suggests that moves to ease monetary policy were fairly effective in sustaining
spending in some policy-sensitive sectors, such as trade, but completely ineffective in
others, notably large-scale real estate, where credit supply proxies appear to be important in determining the level of activity. In other sectors, particularly consumer spending, activity was unusually weak, but credit supply restrictions appear to be less
important.
Ultimately, the paper leaves an empirical puzzle: why was aggregate demand, particularly household spending, so weak overall? While lower expected future income
growth could have produced this weakness, the timing of the spending slowdown suggests that credit factors, particularly balance sheet effects, played a role. Indeed, balance sheet restructuring by consumers and firms (perhaps combined with long-term
changes in financial structure) may be able to account for what econometrically appears
as a large nonlinear shift in the relationship between credit and activity.
Finally, structural changes over the 1980s may have made the credit-output-policy
link impossible to estimate accurately. Because of the large changes in financial structure since the last crunch and recession, historical relationships between credit, activity,
and other financial variables may no longer be valid. Estimating relationships over just
Chart 22: Reduced-Form Prediction Errors Including Mosser and Steindel Residuals
Actual Less Predicted
$1987, Billions of Dollars
25
$1987, Billions of Dollars
40
Producers' Durable Equipment\
Consumer Durables
\
1^
With Short-term
Credit Residuals
$1987, Billions of Dollars
15
$1987, Billions of Dollars
20
Residential Structures
With Business ^v Mortgage Residuals
-40
-60
With Household and
Business Mortgage
Residuals
1988
89
90
Note: Shaded areas represent recession.
295
Chart 23: Reduced-Form Prediction Errors
Actual Less Predicted
$1987, Billions of Dollars
80
Consumer Durables
With Consumer
Fed Funds Spread
s
v
60
$1987, Billions of Dollars
40
Nonresidential Construction
.40
With Hamdani, Rodrigues
& Varvatsoulis Residuals
20
0
-20
-40
-60
1988
90
89
91
92
1988
90
91
92
$1987, Billions of Dollars
40
$1987. Billions of Dollars
25
Original
Producers' Durable Equipment
Producers' Durable Equipment:
With Hamdani, Rodrigues
& Varvatsoulis Residuals
Original
1988
89
90
1988
91
90
89
91
92
Note: Shaded areas represent recession.
Table 6: Summary Prediction Errors with and without Credit Supply Proxies8
296
Sectors
Reduced form prediction
errors 1990-1991
Consumer
Durables
Residential
Structures
Producers'
Durable
Equipment
Nonresidential
Structures
-3.0
-9.0
0.6
-4.3
Prediction errors with credit proxies from:
Lown and Wenninger
-2.3
-8.6
0.5
-4.3
Mosser and Steindel
-4.8
-16.4
0.2
-1.5
Interest rate spreads
10.5
Small firm proxies
a
0.5
0.0
Prediction error (actual-predicted) as a percent of actual spending.
0.1
-
the 1980s may also give inaccurate readings, because much of the transition in financial
structure was in progress during the decade. If this is the case, any measurement of the
"new" credit-output-policy link will only be possible with more data and more time.
Reference
Akhtar, M.A., and K. Harris. "Monetary Policy Influence on the Hconomy: An Empirical Analysis," Federal Reserve Bank of New York Quarterly Review, Winter 1987.
Hirtle, B. "Loan Commitments and the Transmission of Monetary Policy," in Studies
on Financial Changes and the Transmission of Monetary Policy, Federal Reserve
Bank of New York, May 1990.
Bennett, P. "The Influence of Financial Changes on Interest Rates and Monetary Policy:
A Review of Recent Evidence," Federal Reserve Bank of New York Quarterly Review, Summer 1990.
Bernanke, B. "Credit and the Macroeconomy," Federal Reserve Bank of New York
Quarterly Review, 18( I), Spring 1993.
Bernanke, B., and A. Blinder. "Credit, Money and Aggregate Demand " American Economic Review, May 1988, pp 435-39.
Bernanke, B., and J. Campbell. "Is There a Corporate Debt Crisis?" Bmokings Papers
on Economic Activity, 1: 1988.
Bernanke, B., and C. Lown. "The Credit Crunch," Bmokings Papers on Economic Activity, 1992:2, pp 205-39.
Blanchard, O. "Consumption and the Recession of 1990-1991," American Economic
Review Papers and Proceedings, May 1993.
Bosworth, B. "Institutional Change and the Efficacy of Monetary Policy," Bmokings
Papers on Economic Activity, 1:1989, pp. 77-110.
Cantor, R. "Interest Rates: A Panel Study of the Effects of Leverage on Investment and
Employment," in Studies on Financial Changes and the Transmission of Monetary
Policy, Federal Reserve Bank of New York, 1990.
Eckstein, O., and A. Sinai. "The Mechanisms of the Business Cycle in the Postwar
Era," in Robert J. Gordon, ed. The American Business Cycle: Continuity and
Change, Chicago, University of Chicago Press, 1986.
Fazzari, S., R.G. Hubbard, and B. Petersen. "Financing Constraints and Corporate Investment," Bmokings Papers on Economic Activity, 1988:1.
Friedman, B. "Changing Effects of Monetary Policy on Real Economic Activity," Monetary Policy Issues in the 1990s, Federal Reserve Bank of Kansas City, 1989.
Friedman, B., and K. Kuttner. "Money Income, Prices and Interest Rates," American
Economic Review, June 1992.
297
Causes and Consequences
298
Hancock, D., and J. Wilcox. The Effects on Bank Assets of Business Conditions and
Capital Shortfalls, Credit Markets in Transition, Proceedings from the 29th Annual
Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago,
1992.
Jensen, M. and W. Meckling. "Theory of the Firm: Managerial Behavior, Agency
Costs, and Ownership Structure," Journal of Financial Economics, October 1976.
Johnson, R. "The Bank Credit Crumble," Federal Reserve Bank of New York Quarterly
Review, Summer 1991.
Kahn, G. "The Changing Interest Sensitivity of the U.S. Economy," Federal Reserve
Bank of Kansas City Economic Review, November 1989.
Kashyap, A., and J. Stein. "Monetary Policy and Bank Lending," Graduate School of
Business, University of Chicago mimeo, 1992.
Leeper, E. "Consumer Attitudes: King for a Day," Federal Reserve Bank of Atlanta
Economic Review, July/August 1992.
Mauskopf, E. "The Transmission Channels of Monetary Policy: How have They
Changed?" Federal Reserve Bulletin, Board of Governors of the Federal Reserve
System, December 1990.
Mosser, P. "Changes in Monetary Policy Effectiveness: Evidence from Large Macroeconometric Models," Federal Reserve Bank of New York Quarterly Review, Spring
1992.
Peek, J., and E. Rosengren. "The Capital Crunch in New England," Federal Reserve
Bank of Boston, New England Economic Review, May/June 1992, pp. 21 -31.
Owens, R., and S. Schreft. "Identifying Credit Crunches," Federal Reserve Bank of
Richmond Review, 1992.
Ryding, John. "Housing Finance and the Transmission Mechanism of Monetary Policy" in Studies on Financial Change and the Transmission of Monetary Policy, May
1990.
Romer, C , and D. Romer. "New Evidence on the Monetary Transmission Mechanism,"
Brookings Papers on Economic Activity, 1990:1.
Sims, C. "Money, Income and Causality," American Economic Review 1972.
Steindel, C. "Changes in the U.S. Cycle: Shifts in Capital Spending and Balance Sheet
Changes," in Changes in the Business Cycle and Implications for Monetary Policy,
Bank for International Settlements, Basle, April 1993.
Throop, A. "Financial Deregulation, Interest Rates, and the Housing Cycle," Federal
Reserve Bank of San Francisco Economic Review, Summer 1986, pp. 63-78.
. "Consumer Sentiment: Its Causes and Effects," Federal Reserve Bank of
San Francisco Economic Review, 1992.
Wenninger, J., and J. Partlan. "Small Time Deposits and the Recent Weakness in M2,"
Federal Reserve Bank of New York, Quarterly Review, Spring 1992.
Wojnilower, Albert. "The Central Role of Credit Crunches in Recent Financial History," Brookings Papers on Economic Activity, 1980:2.
299
Causes and Consequences
300
The Credit Crunch and the Construction Industry
by Ethan S. Harris, Michael Boldin, and Mark D. Flaherty1
For a number of reasons the construction industry seems like a good place to start an
investigation of the credit crunch. Real Estate was particularly vulnerable to both major
elements of the crunch: the tightening of capital and lending standards, and the "credit
crumble"—the chain of causation running from real estate prices to the capital position
of lenders to their willingness to lend. The decline in construction pre-dated, and seemingly helped set in motion, the general economic decline: while the overall economy began to slow in the spring of 1989 and did not fall into recession until the summer of
1990, construction spending peaked in 1986 and began to decline at an accelerating rate
in 1988. The contraction in real estate was also deeper than for the economy as a whole,
accounting for about one-fourth of the peak-to-trough decline in GDP and a similar portion of the decline in GDP relative to potential from 1989-1 to 1991-I.2
The direct effects of the real estate decline represent only half of the story. Equally
important have been the indirect channels, including the "multiplier effect" of reduced
construction spending on overall economic activity, and the "balance sheet effect" of
crumbling real estate values on the ability and willingness of depositories to lend and
households and businesses to borrow.
Credit crunch proponents can marshal! considerable anecdotal and survey evidence
to support their case. Not only are real estate horror stories featured in most articles that
allege a credit crunch, but a variety of surveys have strongly supported a crunch in real
estate. Not surprisingly, the loudest complaints have come from the borrowers. The
business loan survey by the National Federation of Independent Businesses has shown
a steady increase in the percent of those reporting that loans arc "harder to get." (Chart
1) That survey also shows that construction companies consistently report more difficulties than other companies. In November 1991 the National Association of Home
Builders (NAHB) reported that 70 percent of builders and developers surveyed reported
1
We arc grateful to many colleagues at the bank that provided useful comments and suggestions. Josh Glcason, Paul Ludwig, and Lara Rhamc provided excellent research assistance to this project.
2
From 1990-11 to 1991-1 real GDP fell $106 billion and real construction fell $25.2 billion. Assuming
potential growth of the 2.5 percent (for both GDP and its components), from 1989-1 to 1991-1 GDP fell
$264.7 billion relative to potential and construction fell $65.8 billion relative to its "potential."
301
Causes and Consequences
Chart 1: Survey of Business Borrowers:
Net Harder to Get
Percent
20
302
Construction Companies
15
10
All Companies
1986
87
88
89
90
91
92
Note: The survey is by the National Federation of Independent Businesses. "Net harder to
get" is the percent of respondents reporting that loans are "harder to get" minus the percent
reporting "easier to get." Shaded areas on this and all subsequent charts represent NBER
dated recession unless stated otherwise.
that changes in bank lending operations had caused downward revisions in their plans.3
And in a third survey of builders in January 1992, 82 percent reported restrictions on
normal lending policy.4
Lenders have also reported a toughening of standards. Last March a survey of commercial real estate professionals, including borrowers and lenders, found that 51 percent
of respondents blamed the downturn at least in part on a significant toughening of loan
underwriting rules.5 The Federal Reserve Senior Loan Officer Survey also confirms a
tightening of lending standards. Since January 1990 the survey has asked whether standards for real estate loans have tightened. At the survey's peak, in the first month this
question was asked, 81 percent of respondents said standards had tightened, while only
19 percent said there had been no change. In comparison, only 32 percent reported
tightening for commercial and industrial loans.
Despite this strong anecdotal and circumstantial evidence, this paper will argue that
the credit crunch—defined as an "exogenous" constriction in the supply of credit—primarily affected the timing, rather than the magnitude, of the decline in construction activity. Although the single-family sector has been weaker than economic fundamentals
would suggest, a wide range of tests fail to find support for the credit crunch, suggesting
that other nontraditional factors such as debt retrenchment, pessimistic investor attitudes and the spill-over from the apartment glut explain much of the shortfall. Unlike
3
Campbell and Polmar( 1991).
4
The
I UCLA center for finance and Real I-stale and was reported by Carlton
The survey was conducted by the
(1992).
•"* This Arthur Anderson survey was reported by Kleege (1992).
the classic credit crunches of the past, problems at banks and thrifts have had little impact on the flow of single-family mortgage credit, due to an active secondary market.
Reduced credit flows to builders probably helped shift activity from small to large builders and may have discouraged the development of new building lots, but the impact on
overall construction was not great enough to leave any tell-tale signs of a shortage in the
supply of homes.
The argument for the multi-family and commercial sectors is more complicated, and
strikes at the heart of how we define the credit crunch. In the 1980s a variety of factors
contributed to overbuilding and "bubble" conditions in these sectors, including easy
lending, loose tax policy and excessively optimistic investor expectations. The subsequent deflation of that bubble was the result of a sequence of events, including tax reform, tightening of lending standards, the recession and the concomitant change in
investor expectations. With all of these shocks to the sector, and with the self-reinforcing dynamics of the bubble deflation, it is difficult to assign blame to individual factors.
We will argue, however, that excess capacity was so pervasive in these sectors—covering virtually every geographic region and persisting unabated despite the collapse in
construction—that it is hard to believe credit restraints played much of a role in determining the level of construction. Once the bubble started to deflate, the grim economics
of excess capacity took over and, except in special cases, no new projects were started.
In other words, the credit crunch may have affected the timing and perhaps the speed of
the contraction, but given the pervasiveness of the excess capacity, the adverse tax environment and the weak economy, it appears that a major contraction in activity was inevitable.
One of the challenges of this paper is to find ways to isolate the credit crunch from
an abundance of other explanations for the weakness in activity. For example, broad
credit aggregates give us little clue because they presumably reflect the equilibrium outcome of both supply and demand forces. We address this "identification problem" by
exploring whether credit restraints were powerful enough to overshadow other shocks
and leave their imprint on any of a wide range of indicators. Since virtually none of the
indicators shows evidence of an impact from the crunch and it seems doubtful that an
important secondary factor would not show up in any of these tests, we conclude that
the credit crunch probably played only a modest role in the real estate contraction.
The paper is organized as follows. Part I briefly reviews the scope of the slowdown
in credit flows and construction activity, explores exactly what we mean by a "credit
crunch," and then reviews our methodology for investigating the impact of the crunch.
The following three sections examine the major sectors of real estate: first reviewing the
changing financial conditions and then weighing the evidence of a credit crunch versus
other explanations of the decline in activity. The evidence ranges from simple descriptive statistics to formal tests of econometric models.6
Real Estate and the Credit Crunch
The Real Estate Recession
Before examining the details of the credit crunch it is useful to put the real estate recession in context. If a severe restraint on real estate lending impeded new construction we
would expect at least some of the following: an earlier and deeper recession for real es6
In addition, a companion paper, Boldin (1993), which is not included in this volume, gives a more thorough review of the housing model simulations.
303
Causes and Consequences
tate than the rest of the economy; a drop equal to or larger than past real-estate downcycles; an unusually low share of construction spending in GDP; and perhaps a larger
decline in lending than in construction activity.
Although the national recession began in July 1990 and was proceeded by a slowdown in activity in 1989, the recession in the real estate sector started much earlier and
has been considerably deeper. As Chart 2 shows, nonresidential construction peaked in
late 1985, falling about 16 percent through early 1987, and then staged a modest comeback before plunging a further 17 percent during the July 1990 to March 1991 national
recession. Since the end of the recession the sector has continued to fall, albeit at a more
modest pace, and as of 1992-11 was 39 percent below its peak. Residential construction
peaked in early 1986, and then began to decline as the multi-family component began a
steady slide that continues to this day. Overall residential construction began to fall
more rapidly in 1989 as single family and home improvements began to decline. The
sharp contraction continued during the recession, but the sector has since rebounded to
its 1986-89 trend line.
Obviously, the drop in real estate activity exceeded the drop in overall economic activity, but this pattern is not unusual. Between 1986 and 1991 total private construction
fell 22 percent, with residential construction declining 27 percent, and nonresidential
construction dropping 16.8 percent. These drops are considerably more than the brief
downturns of 1966 and 1969-70, when declines in total private construction were only
9 percent, but less than the longer drops of 1973-75 and 1978-82, which saw declines in
total private construction of 31 and 23 percent, respectively.
While this decline in real estate activity also preceded the latest national recession,
it is noteworthy that the lead time was considerably longer than in the past. Normally,
nonresidential activity peaks several months before a recession, while residential activity leads the economy by one or two years. In the current cycle the lead time for both
sectors was about five years. As we will argue later, however, this apparent long lead is
Chart 2: Construction Put-in-Place
304
$1987, billions
Residential
200 -
\»* ^¥ 1 /
150 -
TAI/
100 -
f Nonresidential:\.
50 -
1964
68
72
76
80
84
88
92
Note: Construction put-in-place data for this and all subsequent charts are smoothed using
a 7-month centered moving average.
largely an artifact of the shock of the Tax Reform Act. Factoring out its effect leaves a
relatively normal cyclical pattern.
Measured as a share of GDP, construction spending looks unusually low (Chart 3).
In fact, in 1991-IV the construction share reached its lowest level in modern history, at
just 5.3 percent of GDP. Part of this weak performance however, was due to a gradual
secular downtrend in the construction share. Relative to its 1964-79 time trend, the
share was 0.7 percentage point below "normal." This is well within the range of past
construction cycles.
Because of the heavy dependence on loans, real estate financing normally closely
shadows construction activity; in the 1980s, however, this relationship broke down.
Chart 4 shows private construction put-in-place, construction loans and mortgage loans,
all deflated by the construction price index. Most prominent is the fact that mortgage
loans also grew faster than activity in the mid-1980s. This rapid growth primarily reflected home mortgage refinancing, especially during 1986-87 and 1989-92 when fixed
rate mortgage interest rates declined sharply. The chart also shows that construction
loans grew faster than activity during the early 1980s, reflecting in part higher loan-tovalue ratios and a decline in the use of alternative financing. In the more recent period,
the gap between loans and activity has returned to a more "normal" range, and the evidence points to a continuation of the current relationship between construction loans and
activity.
With these broad measures conforming to historical norms, can we still argue that
there may have been a credit crunch? The answer is clearly "yes"; there are a number
of reasons to have expected real estate to hold up better in this downturn than in the past.
Previous periods were marked by a number of shocks that are generally absent today:
rising interest rates, deposit disintermediation (in 1966, 1970, 1973-75 and perhaps
1979-82), and a deep recession due to a severe supply shock (in 1973-75 and 1979-82.)
And although it is true that tax reform played a strong role in the current cycle, the same
Chart 3: Real Construction Relative to GDP
Percent Share
u
%
\'
1/
9
\
\
\
\
N
>. •
|
;
~ - J-0.7
* *• - ^ -0.6 /
J
8
.3
V
i
| |
\
's\
%
7 -
1
6
1
1964
68
V\\
|
72
ifi
i
\
\
\\\
1
V
|N
76
80
1
!1964-79 Trendline
:
^
:
\
84
88
\
A
^^*
92
Note: Series is defined as private construction put in place divided by GDP, all in constant
1987 dollars.
305
Causes and Consequences
Chart 4: Real Estate Financing and Activity
$1987, billions
1000
800
600
400
200
1970
76
79
82
85
88
91
Note: The loan data are from the U.S. Department of Housing and Urban Development and
are deflated by the implicit price index for construction put-in-place. The loan data in this and
subsequent charts are smoothed using 7-month centered moving averages.
could be said for the curtailment of Real Estate Investment Trusts (REITs) in 1973-75.
This broad overview of the "crime scene" suggests at least the possibility of something unusual in this real estate cycle. To make a convincing case, we need to look more
closely at the mechanism of the credit crunch and compare it to alternative explanations
for the weakness. This, in turn, requires that we be quite precise about what we are looking for; in other words, we have to define what we mean by a "credit crunch."
The Classic Crunch
306
In theory, a credit crunch can occur any time there is imperfect access to credit markets.
Until recently, however, the term was reserved for periods of thrift deposit disintermediation in 1966, 1969-70, 1973-75 and perhaps 1979-82.7 In the classic credit crunch,
interest rates rose above regulated deposit rates at thrifts, causing an outflow of deposits.
For example, in 1973 and 1974 the gap between the rate on passbook accounts and
three-month treasury bills grew to 300 basis points. As a result, deposit growth slowed
from 17 percent in 1972 to 6 percent in 1974.K Home buyers had few alternative sources
of financing since thrifts comprised almost two-thirds of the mortgage market and the
nearest alternative lenders, commercial banks, were under disintermediation pressure as
7
A number of other episodes have had "credit crunch" qualities. These include: credit controls during the
Korean War; the impact of the Pcnn Central collapse on the commercial paper market in 1970; Herstatt
Bank's disruption of the Euro-dollar market and the shock of Franklin National's near failure to the CD
market in 1974; the near default of NYC and the municipal bond market in 1975-6; and controls on credit
card and other debt in 1980. Of these, only the 1980 credit controls had an impact comparable to the classic housing crunch.
With consumer price inflation of roughly 11 percent in 1974, this implies that in real terms deposits were
falling at about a 5 percent annual rate.
well. Furthermore, because thrifts were already heavily specialized in illiquid mortgage
assets they were unable to fund new mortgages by selling off existing assets.9
Financial deregulation and innovation in the late 1970s effectively eliminated the
classic credit crunch. Ceilings on deposit rates (and lending rates) were gradually raised
and then phased out by 1986. At the same time, because of deregulation and financial
innovation, mortgage markets became increasingly integrated with other capital markets. In particular, portfolio restrictions and tax inducements for thrifts to specialize in
mortgages were eliminated or scaled back, and two major financial innovations made
mortgages a more attractive and widely available investment vehicle—Adjustable Rate
Mortgages (ARMs) and Collateralized Mortgage Obligations (CMOs). ARMs reduced
the risks to lenders by making it easier to match the interest return on their assets to their
liabilities while at the same time giving borrowers a more flexible menu of choices.
CMOs made mortgage financing attractive to a broad spectrum of investors, sharply reducing the role of deposits in funding mortgages.
As a result, in the last period of high interest rates, 1979-82, the credit crunch in
housing was much less severe than in the past. Akhtar and Harris (1986-7) found that
credit rationing had only two-thirds as big as the impact in the earlier periods, Ryding
(1990) found only one-third the effect, and Throop (1986) argued that there was no credit rationing at all. Indeed, this effective end to credit rationing has led several economists to argue that monetary policy changes now have more gradual, less dramatic
effects on the economy.10
A New Kind of Credit Crunch
In contrast to the relatively simple classic credit crunch, in current usage the term has
become a catchall for anything that ails the loan market. Some applications of the term
focus on real estate:
"The real estate markets are in the midst of an unprecedented
credit crunch, brought on by overbuilding in the 1980s."11
But perhaps the most common application covers anything that has caused bank lending
to decline:
"The credit crunch had been caused both by banks' reluctance to
lend and by weak demand from the business community to
borrow."12
Neither of these uses of the term "credit crunch" is very helpful. Overbuilding has
certainly plagued real estate markets but its main effect has been to reduce the demand
for economically justifiable real estate loans. In this sense the credit slowdown is a
symptom of overbuilding and plays no direct causal role in the weakening of the econ-
9
Hcndcrshotl (1991b) lists four factors that made housing markets vulnerable to disintcrmediation in the
1960s and 1970s: (1) interference with the price mechanism: deposit ceilings were imposed to cushion
thrift against adverse interest rate movements; (2) poor asset-liability matching: federally chartered institutions were prohibited from originating adjustable rate mortgages (ARMs); (3) segmented markets: portfolio restrictions and tax inducements caused thrifts to supply two-thirds of the mortgage market; and (4)
inflexible portfolios: regulations and artificially depressed mortgage rates prevented development of a secondary market for mortgages.
10
For example, see the discussion in Mosser (1992).
11
Klccge(1992).
12
Greenhouse (1992).
307
Causes and Consequences
omy.13 The second use of the term equates a decline in the growth of credit with the
credit crunch. By this definition every period of weakness in the economy would be
called a "credit crunch".
This paper adopts the following definition of a credit crunch:
"A contraction in the supply of credit to a group of borrowers who
are creditworthy under previous standards of prudential lending."
It is worth noting that this definition extends beyond the classic credit crunch. In the
classic crunch, credit standards were temporarily tighter than "normal"; our definition
also allows for a crunch when credit standards return to normal following a period of
laxity. In the course of this paper we will argue that a large portion of the recent crunch
has been of the later variety and would probably be more accurately labeled a "credit
standards correction" rather than a "crunch."
The current crunch mechanism is more complicated than the classic crunch and is
therefore harder to detect. The classic crunch involved a simple disintermediation
mechanism, essentially affecting only one sector, over a distinct period of time. The
current crunch operates through a multitude of direct and indirect channels. It evolved
over an extended period, starting as early as the Fall of 1989 (or as late as the Spring of
1990) and ending as early as the Spring of 1991 (or continuing today), and has allegedly
affected virtually every sector of the economy.
The Chronology of the Crunch
308
The real estate literature generally points to the passage of FIRREA, in August 1989, as
the "official" beginning of the credit crunch. FIRREA adopted the capital rules of the
Basle Accord and also put in place the mechanisms for resolving failed thrifts. FIRREA
has several implications for real estate markets:
• Tighter capital requirements encouraged thrifts to try to shrink their assets, including
real estate loans.
• The risk-weighting of assets encouraged banks and thrifts to focus their asset shrinkage on real estate loans. Acquisition, Development and Construction (AD&C) loans
and multi-family and commercial mortgages all were given the highest risk rating,
100 percent, while nonguaranteed single-family loans were given a 50 percent
weight and government guaranteed loans and securities were given a zero weight.
FIRREA chartered the Resolution Trust Corporation (RTC), starting the process of
systematically closing failed thrifts. Although, at least in theory, other lenders
could fill the gap and lend to creditworthy borrowers, presumably they would not
do so on the same generous terms as a failing institution and it would take lime to
develop relationships with the displaced borrowers.
• It limited lending to 15 percent of capital to any one borrower (down from 100 percent of capital). Presumably, this change greatly affected the nonresidential and
multi-family sectors. It may have also hurt some large single-family builders as
well. In addition, nonresidential loans were limited to 400 percent of capital.
• Thrifts were prohibited from taking direct equity stakes in real estate investment, a
relatively common practice in the 1980s.
In 1991, the second major legislation was passed, FDICIA, which established risk'•* To be precise, overbuilding did play an indirect role in the credit crunch to the extent that it has caused
defaults on past loans, impairing bank capital and reducing lending.
based capital rules for banks and mandated that regulators develop stricter guidelines on
real estate lending. FDICIA's main effects on real estate lending were:
• Through a variety of penalties and incentives FDICIA encouraged banks and thrifts
to be "well capitalized" and maintain high risk-weighted capital ratios.
• It required the four federal regulatory agencies to establish real estate lending standards and they adopted the following loan-to-value guidelines a year later: raw land
(65 percent), land development (75), multi-family and nonresidential construction
(80), 1-4 family construction (85), improved property (85), and 1-4 family mortgage and home equity loans (90).
• Subsequent regulations tightened the requirements for professional appraisals.
Although many of these rules were not adopted until late 1992 and a variety of exceptions were allowed, there had been earlier proposals for even tighter requirements,
discouraging some loans in the interim.
Two nonstatutory developments reinforced the impact of the new real estate lending
rules. First, the collapse in real estate values, starting in the oil patch and developing
later in New England and elsewhere, put further pressure on the capital position of lenders. This "capital crunch" was particularly important given that regulators were tightening rather than easing capital requirements. Second, there were significant changes
in attitudes toward real estate loans on the part of borrowers, lenders and regulators.
Early in 1990, in reaction to the large losses at thrifts, there was a general call for tighter
standards from the public, elected officials and some regulators. At the same time, the
adverse experience in some real estate markets (commercial and multi-family) and for
some kind of lenders (thrifts taken over by the RTC) probably caused closer scrutiny of
all kinds of real estate loans. Surveys of borrowers during this period complained of
stricter collateral requirements, withdrawal of lines of credit, stricter evaluation of
projects, tougher occupancy requirements, and of regulators forcing lenders to increase
their loan-loss reserves. Indeed, some commentators point to regulatory pressure on the
Bank of New England to increase reserves for bad real estate loans in late 1989 as an
important trigger-date in the credit crunch.
As the national recession deepened, several steps were taken to ease the credit
crunch. In March 1991 the major bank and thrift regulators issued a joint statement urging continued lending to sound borrowers, while assuring lenders that examiners would
not value loans at depressed liquidation values. Although the statement was followed
with a second clarifying statement in November 1991, these steps did not prevent the
GAO from issuing a letter in March 1992 criticizing regulators for too stringent real estate lending rules. By early 1992, surveys showed the proportion of respondents reporting that loans are harder to get leveling off. More recently, in the spring of 1993 the
regulators announced several modest steps to provide regulatory relief and encourage
lending. The last two years have also seen a steady improvement in the capital position
of lenders. With real estate prices apparently stabilizing it appears that the capital
crunch is over.
Identifying the Crunch
Not only did the credit crunch gradually unfold, rather than suddenly slamming the
economy, it interacted with other shocks to the economy in a more complex fashion than
past crunches. Chart 5 illustrates the complex relationship between credit flows and real
estate activity. The chart shows how seven kinds of shocks to the loan market (the out-
309
Causes and Consequences
Chart 5: Credit Flows and Real Estate Activity
310
Decline in
Aggregate
Demand
Weaker
Economy
Tougher
Capital
Rules
Higher
Vacancy
Rates
\
Lending
Restrictions
Loan
Defaults
Fewer
Loans
Greater
Risk
Aversion
I
Attempt to
Reduce Debt
—I
J*
Lower Asset
r Asset 1 / ^
Prices/Returns
'Returns |
•
Nee(Jto
Reb jild
Cap ital
V
Tax
Reform
Decline in
Investor
Confidence
Demand Shocks
Both Supply and Demand
Supply Shocks
Aggregate Demand
- defense cuts
- initial monetary tightening
- oil price increases
Tax Reform
Decline in Desired Debt/income Ratio
Decline in Investor Confidence
Greater Risk Aversion
Tougher Regulation of Capital
Tougher Regulation of Loans
side boxes) have directly and indirectly caused a decline in both lending and overall economic activity. Although the diagram is rather complex, it simplifies by ignoring some
feedback effects. Clearly two of the shocks—tougher regulation of lending and more
stringent capital requirements (eg. the "capital crunch")—are potential elements of the
credit crunch. Certain other shocks, such as increased risk aversion by lenders, could
be considered part of the credit crunch as well.
It is also clear that many determinants of the credit slowdown are not part of the credit crunch. The most obvious example is the slowdown in lending due to the decline in
aggregate demand. Weak economic activity not only reduces demand for loans, but it
makes some projects that would have been creditworthy in a strong economy uncreditworthy. Reluctance to finance new construction due to the overhang of unused space
and the desire of borrowers to pare down debts should not be considered part of the
crunch. Similarly, although changes in the tax treatment of real estate go a long way in
explaining the weakness, they are conceptually quite different from a credit crunch, and
instead should be considered a determinant of the demand for credit.14
All of these supply and demand factors became entangled once the real estate contraction got under way. This makes it very difficult to measure the indirect or secondary
channels of the credit crunch. For example, one of the principle credit crunch mecha14
The shocks to credit demand—high leverage ratios, excess real estate capacity, tax reform and the recession—all reduced the credilworthincss of borrowers. The resulting contraction in credit should be considered separate from the credit crunch regardless of whether it caused borrowers to stop seeking loans or if it
caused lenders to reject requests for loans.
nisms was the "credit crumble"—the impact of the decline in the price of existing buildings on the capital position, and willingness to lend, of real estate lenders. This fall in
prices reflected all of the negative shocks to the sector, including tax reform, the weak
overall economy, and credit tightening, as well as all of the forces that contributed to the
prior overbuilding. Clearly only part of the "blame" for the resulting pullback in lending
should be apportioned to the credit crunch.
Empirical Methodology
Past credit crunches have been relatively easy to characterize and control for in empirical analysis. Only one sector of the economy was involved, the time period was clearly
delineated and simple proxy variables were available to measure the intensity of the
crunch—the spread between deposit and market interest rates, and the growth in deposits at thrifts were good measures of disintermediation pressures and could be treated as
more or less exogenous variables. Indeed, because each crunch had similar characteristics a simple 0/1 dummy variable could be used to capture the "average" effect across
credit crunches.
Today's much more complicated crunch mechanism does not yield any useful
proxy variable. For example, a 0/1 dummy variable would not be appropriate since
many sectors of the economy are involved, the time frame is not clear and a dummy
variable cannot distinguish between the effects of the crunch and other missing variables. Lacking a "smoking gun" we are forced to resort to examining a variety of circumstantial evidence.
We approach the problem in two steps. First, we explore whether "fundamentals"
can explain the construction decline, using both formal econometric models and descriptive analysis. Presumably any "residual" drop in activity reflects nonfundamental
factors such as the credit crunch. Second, we see how the credit crunch stacks up against
other nonfundamental factors. If a credit crunch constrained the supply of real estate,
we would expect telltale signs in a number of indicators. Did "constrained" lenders cut
back more than "unconstrained" lenders? Did indicators of excess demand, such as vacancy rates, real estate prices, inventories and interest rate spreads, show signs of a
shortage of credit? Individually, none of these indicators gives very compelling evidence about the importance of the credit crunch. Taken together, however, they give a
useful sense of whether the credit crunch was an important secondary factor in the real
estate contraction.
In addition, for the multi-family and nonresidential sectors, we explore whether
problems of excess capacity overwhelmed any effect from the credit crunch. Was the
path of construction in these sectors sustainable for any major market under reasonable
economic assumptions? Did the actual fall in construction put the sector on a sustainable path? Was it likely that new construction projects would have generated a high
enough income stream to cover construction costs? Would the path of construction
spending have been materially different without an exogenous shock to lending?
As we will see, neither of these approaches can fully reject an important role for the
credit crunch, but they do cast a considerable shadow of doubt on the hypothesis.
II. Single-Family Construction
In the past the single-family sector has been the main victim of credit crunches, so it is
a natural place to start our investigation. In the last thirty years there have been five single-family construction cycles (Chart 6). Of course, in each cycle a lot more than a cred-
311
Causes and Consequences
Chart 6: Single-Family Real Estate Financing and Activity
$1987, billions
140
Construction
Put-in-Place
120
100
80
60
40
20
1964
68
72
76
80
Note: Loan data are not available before 1970.
it crunch was happening. We can crudely control for the most important other factor—
the overall economic cycle—by comparing the peak-to-trough movement in singlefamily construction over its own cycles and over the national business cycle. By either
standard the current cycle does not stand out. The current cycle is about average relative
to past construction cycles, and as in past recessions the sector recovered most of its initial decline within the year following the recession. Nonetheless, given the significant
decline in mortgage interest rates and the concomitant improvement in home affordability, both the severity of the decline and the weakness of the recovery in single family
construction have come as a surprise to housing experts.
In this section we first examine the modus operemdi for the current versus past
crunches. We then line up and investigate the various suspects that are allegedly responsible for the housing contraction: demographics, tax reform, debt retrenchment, changes in investor sentiment and the national recession as well as the credit crunch. Finally,
we gather a considerable amount of circumstantial evidence—interest rate spreads, loan
shares, lead-lag relationships, measures of excess demand and model forecasts—all
suggesting a small role for the credit crunch in the decline in residential construction.
A More Complicated Credit Crunch
312
In contrast to the classic credit crunch, the current crunch has not been particularly directed at the single-family sector. The risk based capital rules do not penalize singlefamily lending as much as other real estate lending. The tightening of single-family
lending rules has also been less dramatic, if only because the prior loosening of standards in the 1980s was less obvious. In addition, single-family mortgages have not experienced the dramatic default rates of other real estate loans, so there has been less
pressure on regulators to reign in this sector. Last, and most important, this sector has
benefitted most from financial innovations, with both ARMs and CMOs becoming commonplace.
Nonetheless, the crunch could have had effects on this sector, particularly on the construction side of the market. The adverse experience in related markets (commercial
and multi-family) and for some kinds of lenders (thrifts taken over by the RTC) probably
caused a more cautious attitude toward single family lending by lenders and their regulators. Furthermore, most single-family construction loans have the same risk weights
(100 percent) as other construction loans. Even with the diversification of mortgage
lending and the rapid growth of securitization the sector remains dependent on the
shrinking thrift industry.
Although today's crunch, unlike the classic crunch, does not yield a useful proxy
variable, it does imply some testable hypotheses. We examine three kinds of evidence.
First, we outline the potential causes of the downturn. Second, we explore whether
housing "fundamentals" can explain the housing decline. And third, we investigate
whether the credit crunch left its "fingerprints" on various housing indicators.
Comparisons with Other Explanations
There are seven primary suspects, or at least co-conspirators, in the housing contraction.
Three of these relate to housing "fundamentals": tax reform, demographics and the recession. The other four are more subjective, special factors:
• as with consumption in general, borrowers may have initiated efforts to retrench in
the face of excessive debt accumulation in the 1980s;
• the perceived riskiness of home investment, such that demand for housing is now
driven more by shelter needs, may have increased;
• the demand for homes may have fallen because of the glut of new apartments; and
• credit flows may have been restricted.
Offsetting these seven negative factors has been the drop in interest rates and real home
prices, and the attendant improvement in home affordability.
Model Simulations
One easy way to isolate the subjective factors from the fundamentals is through model
simulations. In particular, we can estimate housing equations over the period before the
housing decline and then use the results in simulations of the alleged credit crunch period. If the models tend to over-predict in the recent period, this will suggest that something unusual was depressing the housing market.
The companion to this paper, Boldin (1993), discusses the results in detail; here we
highlight its findings. Boldin first looked at some existing models and then considered
several variants that attempted to improve on these specifications. Although the models
are constructed quite differently and have varying degrees of econometric "purity," taken as a whole they confirm that fundamentals cannot explain all of the recent weakness
in housing. Overpredictions of 20 percent (over $30 billion) or more at the 1991-1
trough in residential construction were common.15
Noting a variety of econometric problems with the existing models, especially their
15
The models that were examined included Ryding (1990), the 1987-viniagc MPS housing equation, the latest MPS housing equations, Esaki and Wachenheim (1984/5), the Eckstein's (1983) DRI model and the
latest version, and DiPasqualc and Whcaton (1990). Some of these models cover the whole residential
sector, but further analysis shows that virtually all of the undcrprcdiction in the "crunch period" came
from the single-family and "home improvements" components.
313
Causes and Consequences
inadequate efforts to capture long run trends, Boldin developed a new specification for
residential construction that uses recently developed cointegration and error-correction
concepts. It allows for short-run deviations from these fundamentals, but it includes an
error correction mechanism that leads to long-run trend reversion. One advantage of
Boldin's approach is that it can be used to formally test the overbuilding hypothesis. He
developed the new model to empirically encompass the best features of the previous
specifications as well.
Boldin's housing model regresses net residential investment (as a percent of the existing stock) on a credit crunch dummy (that equals one in I969-III to 1970-111 and
1973-1V to 1975-1 and zero elsewhere), changes in nominal mortgage rates, a housing
affordability index (mortgage payments as percent of median household income), the
new home inventory-to-sales ratio, changes in the unemployment rate, and consumption
growth. The model also includes a long-run trend reversion variable, which is estimated
on the assumption that the desired housing stock grows smoothly over time after adjusting for market prices and the cost-of-capital. The underlying theory behind this view is
that utility maximizing homeowners try to maintain the real opportunity cost of housing
as a constant share of their wealth (see Boldin 119931 for more details).
Chart 7 shows the in-sample (it (from 1968 to 1989) and the out-of-sample predictions (1990 to 1992) for Boldin's model as applied to the single-family sector. Although
actual spending fell about $30 billion from 1989-1V to 1991-III, the model predicts a
much smaller $5 billion decline over this period. This $25 billion unexplained drop persists into the housing recovery and is consistent with estimates from the other housing
models.
What causes the overprediction? The chart shows two additional simulations that
can shed light on this overprediction. In the first simulation, we hold both nominal interest rales and mortgage affordability constant after 1990-11. The chart shows that
Chart 7: Single-Family Construction:
Model Simulation Results
3I4
$1987, billion
IDU
In-sample fit
140 -
Basic Model
i
/ -
120
/ /
100
80
t
Actual
(7
/
without
Interest Rate
Effects
60
40
X-
Crunch
Effects "
on
1980 81
82
83
84
85
86
87
88
89
90
91
92
Note: The simulations start in 1990:Q3. For the alternative simulations either the interest
rate effects are turned off or the credit crunch dummy is set to one.
without these short-run effects the sector would have had a modest downcycle in this
period. Comparing this simulation to the base case simulation shows that interest rates
factors should have stimulated housing investment. This counterfactual fall in singlefamily construction is due to increases in unemployment and low consumption growth.
Therefore, the model concludes that falling interest rates should have more than offset
the impact of a weak economy. In the second simulation we allow the same 0/1 dummy
variable used for past crunches to be "turned on" over the 1990-111 to 1992-1V period.
With this change, the model almost exactly predicts the depth of the downturn, even
with lower rates. In other words, whatever is causing the unusual weakness in housing,
it is having an effect that is quantitatively the same as previous credit crunches.
The simulations do not take into account new demographic trends. Since home ownership is closely tied to starting a family one would expect changes in the rate of growth
in the prime new-owner age group, twenty-five to thirty-four, to affect construction. Indeed, the expected decline in this age group is the basis for Mankiw and Weil's (1989)
frequently cited prediction that real home prices would slide about 50 percent over the
next two decades.
After peaking at 1.3 million per year in the late 1970s, the growth in the population
of twenty-live to thirty-four year-olds slowed to 0.7 million in the early 1980s. Since
1986 growth has fallen steadily and by 1992 this group was shrinking at a 0.5 million
annual rate. Unfortunately, at best, demographic variables are not statistically distinguishable from other trend-like variables in housing models. At worst, they were extremely misleading in the 1980s, as shifts in the home-ownership rates within age
groups dominated any effect of changes in the size of different age groups.
Thus, Boldin found that when applied to data that includes all of the 1980s, the demographic variables used by other researchers either had the wrong sign or were quite
small and not significantly different from zero. The most reasonable conclusion is that
supposedly important demographic effects are not very evident in national housing
trends. This claim is confirmed by the fact that (at least so far) Mankiw and Weil's predicted decline in real home prices has turned out to be grossly false. Despite the 199091 bust in the housing market, a very slow recovery out of the recession, unfavorable
demographic trends, and a small decline in home ownership, real home prices have
stayed roughly stable since the mid-1980s. It is also worth noting that housing models
which attempt to exploit demographic trends had similar or worse prediction problems
as Boldin's preferred ECM specification.
The notion that something beyond demographics and other fundamentals is behind
the housing decline is confirmed by private sector housing forecasts during the period.
By the fall of 1989 the effect of adverse demographics and Tax Reform should have
been fully factored into housing forecasts. A typical forecast, such as Data Resource
Inc. (DRI), was assuming modest economic growth and relatively flat mortgage interest
rates. In this environment, they predicted that single family housing starts would remain
flat at about 1.1 million. Actual starts in 1991 were about 200,000 lower, showing about
the same magnitude of overprediction as the econometric models.
Since housing fundamentals do not explain the weakness we are left with at least four
"special factors": a generalized effort to reduce debt, an adverse shift in investor psychology, a spill-over from the glut of apartments, and a credit crunch.
First, home-buyers may have reacted to the debt accumulation of the 1980s and decided to reduce consumption, including consumption of housing services. As Chart 8
shows, the ratio of mortgage debt to the value of single-family homes grew about 15 percentage points in the 1980s. Despite declining mortgage rates, debt service payments
as a share of personal income also rose. Both series turned down in 1991, although the
315
Causes and Consequences
Chart 8: Household Mortgage Debt Ratios
316
Percent
Percent
50
8
Mortgage debt service
payments as a share
of disposable persona^
income
45
40
\>
Households' home <^
; ; mortgages as a share >:\
^
of owner-occupied
^
;
real estate
><
: :
35
(right scale)
30
1960
64
68
72
76
80
84
88
92
25
Note: Mortgage debt service payments data estimated by the Federal Reserve Board and
smoothed by the Federal Reserve Bank of New York. Home mortgages and owner-occupied
real estate data obtained from the Federal Reserve Board's Flow of Funds accounts.
debt ratio has since rebounded. It is plausible to argue that at least some of this retrenchment was due to a conscious debt-reduction effort by borrowers.
Another sign of the debt-aversion of homeowners comes from the data on mortgage
refinancing. With interest rates down sharply, homeowners could cut monthly payments
equally sharply. Instead, many homeowners are opting for shorter term mortgages with
equal or higher monthly payments. For example, according to data from the Mortgage
Bankers Association, about 40 percent of those refinances (in 1992) who had thirtyyear loans opted for fifteen-year loans. This occurred despite a stable spread between
thirty-year and fifteen-year interest rates and despite the prospect that higher future taxes would increase the attractiveness of mortgage interest deductions.
A second nonfundamental factor is the apparent change in attitudes toward home
ownership as a form of investment. As Chart 9 shows, home prices grew at an annual
rate of 2.1 percent above inflation in the 1970s. Although real home prices fell in the
early 1980s it is reasonable to assume that homeowners saw this as an aberration
brought on by severe economic recession and sky-high mortgage rates. The same could
not be said for the more recent period, with equally poor returns despite better housing
fundamentals. As a result, judging from the outpouring of news reports on the perils of
home investment, attitudes appear to have changed dramatically: homes are now seen
as a riskier investment.16
A third possible explanation, which reinforces the second, is that the apartment glut
may have depressed housing demand. As we will show in Part I I I , overbuilding has
caused apartment vacancy rates to reach record levels in the late 1980s. Although reliable data is hard to come by, the residential rent component of the Consumer Price Index
16
Recall that housing models generally focus on home affordabilily—which has improved dramatically—
rather than the return to home investment—which has deteriorated equally dramatically.
Chart 9: Real Quality-adjusted New Home Prices
Index 1987=100
110
105
100
95
Annual Real Home Price Inflation
1969-79 1979-89
U.S.
2.1
-0.8
Northwest
NA
2.3
Midwest
NA
-1.5
South
NA
-1.2
West
NA
1.0
90
85
80
1968 70
74
76
78
80
82
84
86
88
90
92
Note: Data are deflated by the consumer price index, excluding food and energy. Quality
adjusted prices are from the Bureau of the Census.
has fallen about 5 percent in real terms over the last live years. With apartments cheap
and easy to find, home prices uncertain, and debt levels high, the obvious question is:
Why invest in a new home?
Circumstantial Evidence
How do we distinguish between these three demand-side factors and the supply-side impact of the credit crunch? Unfortunately, the simple approach of looking at overall sectoral credit flows is not very useful. Since virtually no one builds or buys a house
without borrowing heavily, it is difficult to distinguish a causal relationship between real
activity and financing in either the Granger (lead-lag) sense or in terms of the amplitude
of the series. All is not lost: we can still examine a broad range of circumstantial evidence. In particular, the credit crunch story has several testable implications; if the credit crunch played an important role in the decline in activity, we would expect to find
evidence of its influence in some important housing indicators.
Hypothesis: Construction loans were more constrained than mortgage loans.
Although there may have been some disruption to mortgage markets, most descriptions of the credit crunch focus on home builders.17 There are several reasons to believe
home-builders have been particularly vulnerable to a decline in the supply of loans.
First, nearly all builders are small- or medium-size companies with limited access to
non-bank financing. Second, in many areas of the country their collateral has been hard
hit due to falling home prices. This is particularly important because over the years an
17
This is a theme of several articles in housing journals. For example, Pol mar (1990) writes: "For the first
time in our history, the producers of housing are facing a serious credit crunch and are having a great deal
of trouble securing financing for land acquisition, land development and construction of new homes
(AD&C loans) while credit generally has remained available to the ultimate buyers of housing."
317
Causes and Consequences
318
increasing portion of construction has been "speculative"—built before a specific buyer
is lined up. Third, there may be a contagion effect from the problems in commercial and
apartment building. If banks and their regulators have "redlined" these activities some
of their ink may have spilled onto single-family homebuilders as well. Finally, builders
have been among the loudest complainers about the credit crunch.
If building loans are being constrained more than purchase loans this should have
caused a shortage of homes for sale. In all recessions weak income growth tends to reduce demand for homes, putting downward pressure on home prices. Credit restraints
put additional pressure on prices: in the classic crunch lack of mortgage credit should
have put further downward pressure on prices, but in the current crunch lack of construction credit should have exerted some reverse upward pressure on prices. Unfortunately,
home price data are not particularly reliable; a better indicator of supply constraints is
the inventory-sales ratio for homes: if the credit restraint was binding, there should have
been an unusually low level of new homes for sale relative to underlying demand; that
is, a low inventory-sales ratio.
Price behavior does not suggest a pervasive housing shortage. In real terms, home
prices have declined steadily since 1987. The cumulative 10 percent drop almost
matches the 1979-82 decline, a period of deep recession and sky-high mortgage rates
(Chart 9). Judging from anecdotal evidence, until recently the market was even more
stagnant than the official price data suggest, with slow turnover and many homes taken
off the market due to lack of demand.
The regional data are also not consistent with a strong supply-side story. Consider
conditions in the Northeast, which is allegedly the most credit constrained region in the
nation. Table I confirms that while the Northeast suffered the largest drop in starts, this
drop was accompanied by the weakest prices, largest rise in vacancies, and highest inventory-to-sales ratio among the four regions. In this region, demand factors appear to
outweigh the credit supply factors. In the Midwest, starts were steady over the 19881991 period and rose to record levels in 1992 as fundamental trends in inventories and
prices were favorable. In the South and West, despite relatively flat prices, starts also
rebounded sharply in 1992 as inventories are at reasonable levels.
Because price and vacancy data are buffeted by several supply and demand forces,
they do not provide very conclusive evidence on the credit crunch; the inventory data
are more convincing. Chart 10 shows that new home inventories have not been unusually tight. From their April 1989 peak, home inventories have fallen less than in either
the 1973-75 or the 1979-82 cycle. More important, the inventory-to-sales ratio has also
fallen less than in past cycles and is well within the range of previous troughs.18
A more recent argument is that lack of development financing has created a shortage
of building lots. For example, a poll by the National Association of Home Builders last
winter found that 38 percent of the respondents saw lot shortages as one of their top
twenty concerns. Lot shortages could not have had much impact on construction during
the alleged peak of the crunch period, however, because in earlier periods no builders
mentioned this as a top-twenty concern. Taken together, these data, as well as anecdotal
evidence, suggest no evidence of a credit crunch during the 1990-91 period.
Hypothesis: The parts of the market that had access to secondary funding
sources should have been less affected than parts without secondary sources.
Some of the most important financial innovations in the last fifteen years have been
in the secondary market for home mortgages. As a result, even if deposit flows slow or
18
Reinforcing this argument is the fact that in previous housing crunches the primary constraint on lending
was mortgages. This should have helped prevent some of the decline in inventories.
if lenders become leery about holding mortgages on their books the flow of funds to borrowers need not slow. Thus, housing experts such as Hendershott (1991 b) argue that the
deposit contraction should have little impact on the market for fixed rate mortgages
(FRMs) that are a small enough ("conform") to be eligible for agency guarantees and
therefore easily securitized. On the other hand, "major disruptions could occur in the
ARM and jumbo FRM markets" since they are still primarily funded by deposits. "This
could lead to reduced housing demand, real prices and home-ownership" (Hendershott
1991b, p. 21) 19
If ARM or jumbo markets were credit constrained we would expect the volume and
terms of these loans to be tight compared to conventional mortgages. The data suggest
otherwise. If the ARM market was constrained, it was not enough to prevent a sharp
drop in the ARM interest rate. As Chart 11 shows the recent twenty-year low in oneyear Treasury bill rates has also pulled down the associated ARM rate to its all time low.
19
A secondary market for jumbos and ARMs has emerged in recent years, but is significantly smaller than
the conforming market in absolute size and in terms of the percent of mortgages passed through.
Table 1: Regional Housing Indicators
1980-85
1986
1987
1988
1989
1990
1991
1992
104.5
227
205
181
131
104
99
112
Vacancy rate (1%)
1.0
0.98
1.23
1.6
1.58
1.6
1.47
1.33
Inventory (months supply)
8.7
7.2
10.12
13.6
15.1
15.8
14.9
10.53
Price ( 1 % change)
7.4
14.7
13.6
2.4
3.0
-2.1
-3.3
ND
113.0
187
203
194
191
194
192
241
Vacancy rate (1%)
1.6
1.5
1.37
1.2
1.43
1.3
1.28
1.23
Inventory (months supply)
7.3
4.59
4.76
5.34
5.08
6.04
5.56
4.01
Price ( 1 % change)
3.2
5.6
6.7
4.4
2.3
0.8
2.7
ND
385.3
503
485
442
410
372
353
451
Vacancy rate (1%)
1.7
2.12
2
1.93
2.25
2.12
2.2
1.77
Inventory (months supply)
6.8
6.18
6.78
6.21
6.23
6.37
5.72
4.91
Price ( 1 % change)
4.0
3.0
3.5
2.3
2.8
0.2
1.9
ND
163.0
261
255
264
272
227
198
253
Vacancy rate (1%)
1.8
1.75
1.82
1.57
1.6
1.75
1.65
1.93
Inventory (months supply)
7.4
4.84
5.51
4.76
5.21
8.43
7.63
5.39
Price ( 1 % change)
2.8
2.5
5.3
5.4
6.3
4.8
0.8
ND
Northeast
Starts (1000's)
Midwest
Starts (1000's)
South
*
Starts (1000's)
West
Starts (1000's)
319
Causes and Consequences
Chart 10: New Homes Sales and Inventories
Thousands
500
Quarters Supply
3.5
Inventory-Sales
Ratio
(right scale)
3.0
400
2.5
300
V
200
2.0
100
1.5
1.0
1964
Note: Sales are measured at a quarterly rate, inventories and the inventory-sales ratio are
the end of quarter numbers.
Chart 11: Comparison of the ARM and One-year T-bill rates
320
^—•/*/**V*V,
1984
85
86
87
ARM/T-bill Spread
88
89
90
91
92
Although the gap between the ARM rate and the Treasury bill rate has widened, a comparison with 1986 suggests that this is a normal response to unusually low rates.
While Chart 12 shows a steep decline in the ARM share of mortgage origination in
1989, this development is not necessarily evidence of a supply restraint. The econometric model in Brueckner and Follain (1989) shows that the increase in the FRM-ARM interest rate spread should have increased the share of ARMs, but that this is more than
offset by the effect of the general decline in interest rates (Chart 13). Our back of the
envelope calculations show that the current ARM share of 20 percent is roughly what
the Brueckner-Follain model predicts.
This leaves only the "jumbo" market as potentially constrained market. Jumbo mortgages make up less than 20 percent of the value of loans for new homes, with a heavy
concentration in the high-priced northeast and western regions. Although it is possible
that credit terms have tightened for jumbos, housing experts have reported only modest
pressure on the spread between jumbo and conforming interest rates (1/4 to 1/2 percentage points).
Hypothesis: Depository lenders were more constrained than nondepositories.
As we argued earlier, the patterns of aggregate home lending do not give useful clues
about the credit crunch. Although the data are poor, the composition of lending does
provide another test for the credit crunch.20
Differences in lending patterns between depositories and nondepositories do provide
circumstantial evidence for the credit crunch story. Chart 14 shows that the share of
mortgage loans by depositories declined significantly after 1989. Similarly, in the classic 1974-75 crunch mortgage loans fell much faster for depositories than nondepositories, driving down the depository share. In the current period, however, the decline in
the depository share primarily reflects the explosion in refinancing through nondeposi20
T h e detailed loan data used in this paper come from the HUD Survey of Mortgage Lending Activity.
These data come from multiple sources, lack conformity in definitions and suffer from double counting.
Unfortunately, they are also the only source of this kind of detailed information.
Chart 12: ARM Share of Mortgage Originations
Percent
80
20 10
u
1984
85
86
87
88
89
90
91
321
Causes and Consequences
tories rather than a constraint on depository lending. Chart 14 shows that depositories
did not lose market share in the market where we would expect crunch effects—the construction loan market. Indeed, nondepositories have been much more aggressive in curtailing construction loans, pushing up the depository share to close to 100 percent.21
21
Lending from nondepositorics may have been constrained by some of the same problems that plagued
depositories. Here we are testing if there was a stronger constraint on depository lending, as the credit
crunch story predicts.
Chart 13: Comparison of ARM and FRM Interest Rates
Percent
16
14
Fixed
Mortgage Rate
\
Mortgage Rate
Fixed/Adjustable;;;
Rate Spread
1984
85
86
87
88
89
90
91
92
Chart 14: Single-family Depositories' Share
Percent
11U
100
sN\:>
^yf**^
Construction
90
80
70
60
Vy^^;:y
Mortgage Loans
\
50
Ad
r
i
1970 72
322
is V
i
i
i
i i
74
76
78
80
82
84
i
i
i
i
86
88
90
92
Hypothesis: Some generalized "price rationing" may have occurred.
With unregulated interest rates it is likely that more of the shock to the sector will be
transmitted via interest rates rather than "non-price" rationing. Thus if banks were more
reluctant to make mortgage loans this should cause an unusual spread between mortgage
and other interest rates. Chart 15 shows two spreads that would be affected by a reluctance to lend. The top panel shows that over the last twenty years mortgage rates have
averaged 174 basis points higher than comparable Treasury Bonds. This gap reflects the
differential credit and prepayment risk of the two investments, as well as the relative
availability of funds in the two markets. Since the onset of the crunch this gap has fallen
below it historic average, suggesting that banks are not trying to discourage mortgage
lending. Indeed, the declining gap probably reflects the steady increase in liquidity in
the market as new financial products are developed.
Chart 15: The Interest Rate Spreads for Fixed Rate Mortgages
1972
74
76
78
80
82
84
86
88
90
92
84
86
88
90
92
Fixed Rate Mortgages and Federal Funds
1972
74
76
78
80
82
323
Causes and Consequences
The bottom panel of Chart 15, which plots the gap between mortgage rates and the
Federal funds rate, tells a different story. This rough measure of the relative cost of
funding investments has reached record levels. During past periods of Fed easing the
gap has grown to an average of about 400 basis points. It is possible that part of the remaining 100 basis point spread reflects reluctance to lend. In simulations with standard
housing models, this "excess spread" could account for a 5 1/2 percent or $5 billion
shortfall in real single-family construction over the 1990-92 period.
Summary: Single Family
There is ample reason to suspect a credit crunch in single-family housing. Depositories
have been under pressure to tighten up their real estate lending. Borrowers, particularly
builders, have complained loudly about the lack of credit. And the usual array of suspects—demographics, the business cycle and tax law changes—do not appear sufficient
to explain a housing recession in the face of declining mortgage rates. Indeed housing
model simulations show a $20 to $30 billion unexplained decline in single family construction at the trough of the recession.
A careful sweep of the "crime scene," however, yields virtually no evidence of a
crunch. Inventories are not tight; prices and other excess demand indicators suggest no
shortage of housing; loans have fallen equally for all kinds of lenders; ARM and jumbo
markets, with little access to secondary financing, appear to be functioning normally;
and the spread between mortgage and Treasury interest rates is low. With little evidence
of a tangible impact from the crunch, and several plausible alternative explanations for
the contraction, it is difficult to put much blame on the credit crunch.
Multi-Family Residential Construction
On the surface, the multi-family sector looks like a clear-cut victim of the credit crunch.
Since its peak in 1986, multi-family construction has been the weakest sector of the
economy, with the sharpest contraction in both output and lending. Moreover, it has
been touched by all of the elements of the credit crunch—the lightening of lending and
capital rules, the implosion of the thrift industry and the credit crumble.
Despite these appearances we will argue that the credit crunch was "dominated" by
a more fundamental problem in the sector: the speculative boom and the resulting overcapacity in the sector. Starting in the late 1980s, a combination of tax reform, weak demographic and income fundamentals and a shift toward more realistic investor expectations made only subsidized or specialized projects economically feasible in most
markets. This left only a limited role for the credit crunch: it constrained some specialized projects; it may have affected the timing of the contraction; it probably hastened
the fall in the price of existing structures, hurting the capital of lenders and thereby constricting loans to other sectors; and if the crunch persists, it could constrain a rebound in
the sector. Nonetheless, given the adverse economic fundamentals and the grim reality
of over-capacity, it is unlikely that the contraction would have looked significantly different in the absence of the credit crunch.
This section is organized as follows. First, we briefly review the multi-family cycle.
We then describe the credit crunch and its interaction with the speculative bubble (later
we will apply this argument to several components of nonresidential construction).
Next, we explore two kinds of evidence. As for the single-family sector, we look for
circumstantial evidence of a particular kind of supply-side constraint. We then document the extent of the overhang of unused space in these markets and, with some coun-
324
terfactual exercises, we explore whether the contraction in construction was a
"reasonable" response to this overhang.
The Multi-Family Cycle
A broad look at the multi-family data suggests that although the sector has been unusually weak during the period of the alleged credit crunch, this weakness could not have
been important to the overall economy. Even at its peak in 1986 multi-family construction comprised less than one-fifth of residential construction and only about 0.7 percent
of GDP. Furthermore, more than half of the subsequent drop in multi-family construction had already occurred by 1989, before the key elements of the credit crunch—the
credit crumble, and regulatory restraints—had begun (Chart 16). Thus even a complete
disappearance of this sector could have accounted for only a small part of the 2.2 percent
drop in GDP in the 1990-91 recession.
Although its macroeconomic significance is probably small, the data do suggest an
unusually large and sustained drop in multi-family activity. At SI0.2 billion in 1992-IV,
real multi-family construction is less than half of its 1989-11 level, and at the lowest level
since data have been compiled (1964). The drop is also more than twice the 32 percent
decline in single-family construction over its down-cycle (1987-1V to 1991-1). Over the
1986-91 period the decline in multi-family activity was greater than in three of the last
four multi-family recessions. Clearly, however small its role in the overall activity,
something caused an unusually steep drop in multi-family activity.
Alternative Explanations for the Decline
A variety of explanations for the multi-family collapse have been put forth, including
adverse demographics, weak overall economic activity, tax law changes and the credit
crunch.
Several of these explanations can be dismissed as only minor contributors. First,
Chart 16: Multi-family Financing and Activity
$1987, billions
70
Mortgage Loans
60
Construction
50 40
30
20
10
0
-
/
s\
1
1970
72
1
'• '• s \
Construction Loans
74
76
78
80
82
84
86
1
1
88
90
v
1
92
325
Causes and Consequences
there is a role for demographics in the multi-family contraction. As many analysts have
pointed out, growth in the prime age group for new apartments—ages twenty-five to
thirty-four—has slowed dramatically, from 3.2 percent in the 1970s to 0.6 percent in the
1980s. In the 1990s, the twenty-five to thirty-four year-old group is expected to decline
at a 1.1 percent pace. However, other important renter age groups, such as the sixty-five
to seventy-four cohort are growing in size. Combining population data with age-based
rentership rates from the Bureau of the Census1 American Housing Survey, we computed a modest slowing in the growth of the demographic demand for apartments (based
on a weighted average calculation), from 2.1 percent in the 1970s to 1.3 percent in the
1980s and a 0.6 projected growth rate in 1990s. It is hard to imagine that these gradually
evolving demographic variables could cause the rapid shifts in apartment construction
that have been observed in recent years.
Neither does the recession take us very far in explaining the drop. Table 2 shows the
change in real multi-family construction from peak to trough of the business cycle and
the sector cycle. As the top of the exhibit shows, recessions have historically had mild
effects on multi-family housing. Only in the 1973-75 recession, when changes in the
tax treatment of Real Estate Investment Trusts crippled multi-family construction, have
we seen declines of major macroeconomic significance. Only about one-sixth of the de-
Table 2: Peak-to-Trough Movements in Multi-Family Construction8
326
Business Cycles
Percent Change
Change in $1987
1969-1V to 1970-1V
3.1
1.0
1973-1V to 1975-1
-58.8
-27.1
1980-1 to 1980-111
-22.7
-5.7
1981-111 to 1982-IV
-8.6
-1.7
1990-111 to 1991-1
-9.0
-1.5
Percent Change
Change in $1987
1966-1 to 1967-1
-35.1
-8.4
1969-11 to 1970-11
-11.2
-3.6
1973-1 to 1976-1
-75.6
-38.9
1979-IV to 1982-11
-33.1
-8.6
1986-11 to 1992-1V
-69.8
-23.6
1989-11 to 1992-IV
-52.1
-11.1
Sector Cycles
a
For "Business Cycles" the peak and trough dates are from NBER; for "Sector Cycles" the
dates are chosen to match the peak and trough in multifamily construction.
cline in construction during the current sectoral cycle occurred during the recession.
Apparently, apartment demand holds up well in periods of weak income growth because
rentership is an "inferior good" relative to single-family housing (and like all inferior
goods, demand for it typically increases during recessions).
That leaves only two plausible explanations for the bulk of the decline: reduced
credit availability (the credit crunch) or a decline in demand for credit due to an overhang of unused space.
The Credit Crunch and the Real Estate Bubble
The credit crunch mechanism was more complicated in the multi-family (and commercial) sector than in the single-family sector. This added complication comes from the
credit crumble—the impact of declining real estate values on the capital position of
lenders and hence their willingness to lend. The credit crumble is intertwined with all
the factors that led to the speculative bubble and bust in these sectors.
Any explanation for the collapse in multi-family construction and the role of the
credit crunch in that collapse must start with the building boom that preceded it. In general, three kinds of developments were behind this boom: changes in tax policy, changes
in lending behavior and "animal spirits"—a catch-all category for anything that contributes to long cycles in this sector.
The key tax change was the Economic Recovery Tax Act of 1981. ERTA significantly shortened and accelerated depreciation schedules on buildings and added other tax inducements for structures investment. As a result, investors were willing to accept low
or negative pre-tax returns on real estate projects, leading to a proliferation of limited
partnerships and syndicates. ERTA also encouraged some "churning" of properties by
allowing investors to resell and depreciate the same property several times.22
The key regulatory change was the Garn-St. Germain Act of 1982. Garn-St. Germain was the outgrowth of problems inherited from the 1970s. Because thrifts traditionally fund long-term assets (mortgages) with short-term liabilities (savings accounts),
rising interest rates in the 1970s ate into their capital. In 1975 insured Savings and
Loans had a ratio of net worth to assets of 5.8 percent, by 1982 this ratio had fallen to
just 3.4 percent and one-sixth of S&Ls had a ratio below 2 percent. To shore up thrifts
the law allowed greater participation in the growing real estate loan market by removing
rules against certain types of lending, allowing higher loan-to-value ratios, and increasing to 40 percent the proportion of assets that may be invested in commercial real estate.
Garn-St. Germain and the Monetary Control Act of 1980 liberalized the rules on rates
and types of deposits allowed at depository institutions. At the same time, regulators
adopted more lenient accounting rules, allowing some thrifts to remain open despite net
worth that by conventional measures was near zero or negative.
There were other related forces that helped build on this boom atmosphere.23 First,
competitive pressures in traditional loan markets forced depositories to either accept lower returns, downsize their operations, or expand into nontraditional areas such as more
speculative real estate lending. Second, low effective capital requirements, deposit insurance, and the perception that many institutions are "too big to fail" created "moral hazard" problems, encouraging banks to invest in high-return, high-risk markets such as
22
Sec Corker, Evans, and Kenward (1989).
23
This discussion draws on Litan (1992), Hester (1992), Browne and Case (1992) and Gourkyo and Voit
(1993).
327
Causes and Consequences
"hot" apartment and commercial real estate markets.24 The incentive for risk taking was
particularly strong at those thrifts that remained open despite low or negative net worth.25
Not only was there a willingness to take risk, there was also a tendency to underestimate the magnitude of the risks in real estate investment. A number of academic and
professional studies during the period showed that over the period for which data were
available real estate had been a consistently high return investment. And, like any speculative bubble the boom tended to feed on itself as investors scrabbled to buy up real
estate regardless of economic fundamentals.
Exacerbating the excessive optimism of builders and their financiers was the long
gestation period of large construction projects. Since it takes several years to plan and
build a new building, once the sector gains momentum it takes time to put on the brakes.
Many analysts argue that this helps explain the continued high pace of construction in
the several years following the Tax Reform Act of 1986. It also implies that at least
some of the sharp decline after 1989 was a delayed reaction to changing conditions in
the mid-1980s.
Deflating the Bubble
328
Credit crunch advocates argue that credit restraints played a key role in the deflation of
the bubble. They note that the tightening of lending rules for multi-family (and commercial) real estate was particularly dramatic, more than reversing the loose standards
of the 1980s. They argue that FIRREA and FDICIA were particularly punishing to
multi-family lending, giving loans to this sector 100 percent weights in the capital standards, limiting loans to any one developer to 15 percent of unimpaired capital, raising
required loan-to-value ratios, forcing banks and thrifts to increase loan loss reserves,
and prohibiting thrifts from taking direct stakes in real estate investment.
On top of the legislative changes, credit crunch advocates argue that public reaction
to the excess of the thrift industry squeezed multi-family investment from two sides: on
the one hand, the resolution of failed thrifts steadily removed generous lenders from the
market; on the other hand regulators put increasing pressure on the remaining institutions to tighten their credit standards.
Low liquidity in the secondary markets for both mortgage loans and existing structures meant that these pressures on lenders were not easily passed off. Outside of singlefamily mortgages, real estate loans have been very difficult to package. Although there
has been some secondary market activity, it is more an indication of market distress than
market liquidity.
Of course credit constraints were not the only source of trouble. The demand for new
apartment buildings was affected by four developments:
• First, adverse demographics—the aging of the baby boom—reduced the underlying
demand for apartments.
• wSecond, the Tax Reform Act of 1986 had a chilling effect on the incentive to invest
in large real estate projects. It removed the generous depreciation allowances for
24
The Depository Institutions Deregulation and Monetary Control Act (DIDMCA) of 1980 added to this
moral hazard problem by raising the deposit insurance ceiling from S4(),(X)() to $ 1 ()().()()().
25
Some commentators argue that there was a "lemming problem": as more investors delved into the market
it became easier to justify their heavy exposure in real estate markets to their stock holders and regulators.
Furthermore, if many financial institutions were heavily involved in the market private risk and systemicrisk became increasingly connected, increasing the prospect of a government help in the event of trouble.
structures and disallowed the tax loophole of using of "passive" losses from real estate investments as an offset to regular income. This effectively ended the widespread use of syndications to finance projects.
• Third, the national recession and the weak growth that both preceded and followed
the recession reduced demand for new apartments and lowered expectations of future demand.
• Fourth, the oil price collapse in 1986 and the stock market crash in 1987 not only
weakened demand for apartments in two key markets (the oil patch and New York
City) but alerted investors to the dangers of speculative investments.
Together these adverse shocks from both the supply and demand side helped set in
motion the reversal of the real estate bubble. Once the deflation got under way it fed on
itself. Falling prices meant that investors could no longer count on capital gains to make
up for any shortfall in income flows on their real estate investments. This, in turn, reduced the demand for real estate putting further downward pressure on prices. Similar
logic applies to the supply side of the market. Falling prices cut into the capital cushion
of lenders, inhibiting new lending and adding further downward pressure on prices.
Chart 17 uses stylized supply and demand analysis to illustrate three possible interpretations of a typical market for new multi-family structures. (The same analysis will
be applied later to commercial structures.) The demand curve is determined by the expected after-tax return to new buildings, which in turn depends on the underlying demand for apartments, the state of the overall economy, the vacancy rate and the tax
treatment of investment. The supply curve is determined by the cost of construction and
the cost and availability of financing. Note that the supply of new buildings is zero if
the rents fall below a level necessary to cover construction costs.
In the recent period, the supply curve shifts up and to the left (to S I ) due to the credit
crunch. Three interpretations of the demand shift are possible:
(5) For a modest shift in demand ( D l ) , the shocks to supply and demand together contribute to the decline in activity in roughly equal proportion, putting little pressure
on prices or other measures of excess demand. This would be an extreme view of
credit crunch advocates.
(6) A more dramatic shift in the demand curve (D2) leaves a modest role for the credit
crunch. There is a sharp contraction in prices and activity falls close to zero.
Without the credit crunch there would have been Q-II of new construction. This is
essentially what we test for in looking at indicators of excess demand.
(7) In the third case the demand curve (D3) shifts below the original supply curve
(SO). Construction continues on projects with significant sunk costs, but only in
special cases are new buildings started. Except for these special cases the credit
crunch has no influence on the level of activity, although it could constrain activity if the demand curve rebounded. This represents the extreme argument against
a binding credit crunch effect.
As we will show below, depending on the particular market a combination of (2) and (3)
occurred.
Weighing the Evidence
Although the data are not as rich in the single-family sector, a variety of circumstantial
evidence can be examined to distinguish among the three cases. In particular, if supply
restraints were a major factor (that is, Case (1) above) in the construction contraction we
329
Causes and Consequences
Chart 17: The Market for New Buildings
330
Rent
Q,
Q,
Space
Rent
UL
Crunch Effect
Space
Demand Effect
Note: The second chart gives a more sophisticated interpretation of the market. It assumes
that demand for specialized buildings is highly inelastic because it cannot be satisfied by
existing structures. The chart assumes the supply curve becomes increasingly inelastic at
high levels of construction as lenders worry about excessive exposure to the market. The
supply curve shifts up only moderately for the highly creditworthy specialized borrowers.
would expect at least some downward pressure on vacancy rates, some upward pressure
on rents and the prices of apartment buildings, and an easing in loan foreclosures and
delinquencies as the surplus of apartments disappears. Furthermore, even if the credit
crunch is only a modest factor in the contraction (Case (2) above) we would expect
"constrained" lenders to cut back more on loans than "unconstrained" lenders. On the
other hand, if the overhang of unused space is the dominant cause of distress in the sector we would expect sharp declines in loans from all sources, continued high vacancy
rates despite falling real rents and low levels of construction, falling building prices, and
persistently high loan foreclosures and delinquencies.
Multi-family loan activity
Although the credit crumble has presumably affected all major multi-family lenders,
most of the regulatory and legal aspects of the credit crunch have been directed at depositories.26 As in the single-family sector, this gives a lever for testing two key implications of the credit crunch hypothesis: (I) the credit crunch should have forced
depositories to cut back lending more dramatically than nondepositories and (2) it
should have induced all lenders to reduce loans relative to construction activity—in other words, force builders to put up more of a stake in their projects.
Although, as we noted earlier, the data are quite poor, they generally are not consistent with the crunch hypothesis. Chart 16 compares construction activity to construction
loans, both expressed in real 1987 dollars. Until the mid-1980s construction loans were
consistently below construction activity, reflecting in large part the role of direct (equity) financing of projects (as well as some coverage differences in the series). In 1985
financing surged ahead of activity and has remained ahead thereafter. During the alleged crunch period, there was some narrowing of the gap, but this may reflect the fact
that in a weak market the most speculative, highly leveraged, projects tend to be cut first.
In other words, despite the efforts of regulators and lenders to tighten loan-to-value ratios, the flow of loans remained high relative to construction activity.
Chart 18 plots the long-run pattern of the depository share of multi-family loans. In
the early 1980s a combination of economic recovery and easy tax and lending rules
caused a surge in both mortgage and construction loans, especially by depositories.
Since the mid-1980s loans have dropped across the board but the share data show no
evidence of a differential constraint on depository lending. In the mortgage market the
depository share has actually risen during the alleged crunch period and is now above
its historical highs. In the construction loan market the depository share was roughly
flat during the worst of the crunch (1990-91) and today remains in the high end of historical experience.
As we noted earlier, these data are of dubious quality, so it is important to carefully
interpret these findings. In particular, experts note that there may be double counting of
construction loans made by commercial banks and their mortgage bank affiliates. Furthermore, some of the increased share of loans at depositories may be unwanted as they
are forced to either roll-over loans or push the borrower to bankruptcy. Yet the double
counting should work in favor of the credit crunch hypothesis since it should mean a
sharper contraction in lending as loans counted twice are no longer made. Whether the
loans are wanted or not, the data do suggest a continued flow of funds to the sector.
26
For example, risk-based capital rules were not adopted by the National Association of Insurance Commissioners until December 1992 and were not slated to take effect until a year later.
331
Causes and Consequences
Chart 18: Multi-family Depositories' Share
Percent
100
1970 72
74
76
80
82
84
86
88
90
92
The Overhang
332
The argument against a strong role for the credit crunch becomes more persuasive if we
look at measures of excess capacity in the multi-family sector. Chart 19 shows how severe and persistent the excess capacity was despite the sharp drop in construction. Starts
rebounded from the 1982 recession and despite a steady climb in the vacancy rate starts
averaged 650,000 per year from 1983 to 1986. Starts tumbled to 300,000 in 1989, and
the vacancy rate began to trend down, suggesting that the sector was moving toward a
sustainable medium-term path. With the recession, however, the vacancy rate again began to climb, despite a virtual halt to construction.
A comparison of vacancy rates across building size gives an idea of the important
role for tax incentives in the overbuilding. Most of this excess capacity was in the larger
buildings most favored by tax shelters. Data from the American Housing Survey shows
a downward trend in the vacancy rate for single-family rental units in the 1980s. By
contrast, vacancy rates for two-to-four-family units and particularly larger buildings hit
unheard of heights. Indeed, after hovering around 4 percent from 1970 to 1983, the vacancy rate for multi-family rentals more than doubled by 1987 and has eased back only
slightly since then.
The regional data show the broad geographical spread of the over-building (Chart
20). In the South, already high vacancy rates combined with an oil-price induced collapse in the local economy pushed vacancy rates to record heights, while the national
recession interrupted a gradual return to long-run equilibrium. In the West and Midwest, judging from the timing of the construction slowdown, tax reform helped halt a
more modest overbuilding. In the Northeast, a severe regional recession prevented any
improvement in the supply-demand balance as construction hit record lows. Indeed,
had there been no construction in the Northeast in 1991 the vacancy rate would still have
matched the record rate of the previous year.
Chart 19: Multi-Family Housing Starts and Vacancy Rate
Million units
1.2
Percent
10
M Housing Starts
A A (left scale)
1.0
0.8
-^hlA 1/
Aii
li
^
<
$^
| |
- 9
Vacancy Rate ;
(right scale) N
$i
- 8
\ iwv \ %
AA
0.6
- 7
0-4 -
\ I-\
0.2 -
^
JWf I /
1
^
^T
¥ \ ^
^ ^
#
W\
>1
S«J^
.
6
VLii. - 5
\
0.0
1964
68
72
76
80
84
88
92
Note: The vacancy rate data are from the Bureau of the Census, "Housing Vacancy Survey."
Other excess-supply indicators tell a similar story. After turning up in the early
1980s, real rents (deflated by the core GPI) have trended down since 1986. By 1992
nominal rent inflation had fallen to just 2 1/2 percent, its lowest rate since 1968.27 Mortgage delinquency and foreclosure rates tell a similar story. Both rose sharply in 1986 in
response to tax reform and the oil shock in the Southwest. Both have also trended down
since then, although they remain high by historical standards.
Model Simulations
One way to illustrate the dominant role of over-supply vis-a-vis the credit crunch is
through a simple simulation model. The model compares the supply of apartments—
determined by new construction and the rate of depreciation of old structures—to the
demand for new apartments—determined by demographic trends and trends in average
space use—to plot alternative paths for the vacancy rate. (The appendix provides details.) For our out-year projections we assume the depreciation rate remains unchanged
and the absorption rate grows in line with reasonable demographic trends. The model
is based on data for large apartment buildings, the most overbuilt sector. Table 3 shows
the results of three such simulations.28
The table underscores the sheer unsustainability of the construction boom. On the
one hand, holding construction near its 1992 trough and using realistic demand projec27
Actual rent inflation may have been even weaker because the CPI rent docs not fully capture the impact of
"teasers" such as rent-free months for new tenants. According to the NAHB, rents for new units fell 5.3
percent from 1991-11 to 1992-11, compared to a rise of 3.1 percent for existing units. Although part of the
weakness in new rents reflects a composition shift towards cheaper, publicly financed housing, at least
some of the drop probably stems from the excess capacity. Finally, since 1988 the BLS has adjusted the
rent index for depreciation, adding about 1/2 percentage point to reported rent inflation.
28 The appendix provides details on the model and the baseline projections.
333
Causes and Consequences
tions results in declining vacancy rates, but the projected level still exceeds 6.0 percent
in 1997. On the other hand, assuming a counterfactual where construction holds steady
in the 1988-97 period (at the 1983-87 average) would lead to a continually climbing vacancy rate even if demand grew at a robust pace of 1.4 percent per year from 1990-97.
The table also illustrates the implausibility of a truly constraining credit crunch effect. Suppose that from 1989 to 1992 the credit crunch reduced apartment construction
by $4.0 billion per year. This is about one-fourth of the total peak-to-trough drop in activity for this sector. Further, suppose the absorption and depreciation rates followed
their historical patterns. This implies that without the crunch, builders would have continued to find financing despite a 1992 vacancy rate of 12.5 percent. This scenario
seems quite implausible when judged by historical patterns. Even in the 1980s, a booming economy, generous tax breaks, and robust apartments absorption could not support
vacancy rates at these levels. In other words, it is hard to imagine that lenders could be
induced to ignore the warning signal from rising vacancy rates in the late 1980s.
Looking ahead, even the current depressed state of construction will leave the sector
overbuilt for years to come. Assuming that apartments are absorbed in-line with demographics and that the 1992 pace of construction continues indefinitely, the vacancy rate
will not reach 5 percent until after the turn of the century. Given this dismal prospect,
it is easy to imagine an even lower rate of construction during the current period.
Chart 20: Apartment Activity by Region
Percent
18
Thousands
250
Northeast
Percent
8.0
Thousands
400
Vacancy Rate
200
150
1
1
1
1
1
1
1
1
1
1
1981 82
83
84
85
86
87
88
89
90
91
ou
u3
92
100
1981 82
85
86
87
88
89
90
91
83
84
85
86
87
88
89
90
91
92
92
Percent
8
12 1000
ou
334
84
Percent Thousands
Thousands
1400
1981 82
83
1
1
1
1
1
1
1
1
1
1
1981 82
83
84
85
86
87
88
89
90
91
u3
92
Table 3: Multi-Family Sector Vacancy Model Simulations
Base Case and Realistic Forecast
Demand Growth
Absorption
Construction
Vacancy Rate
1977
1.7
17.5
16.1
3.5
1978
1.8
18.1
18.4
3.5
1979
2.3
20.9
21.3
3.5
1980
2.0
20.1
19.3
3.3
1981
1.3
17.4
18.9
3.6
1982
0.9
16.0
17.4
3.8
1983
2.1
21.5
23.3
4.1
1984
1.1
17.4
28.6
6.2
1985
1.4
19.5
31.5
8.3
1986
3.3
28.9
34.0
8.9
1987
1.8
22.9
27.5
9.5
1988
2.4
26.2
24.7
9.1
1989
1.7
24.1
22.3
8.7
1990
1.0
20.5
19.7
8.4
1991
0.2
16.9
15.7
8.2
1992
-2.5
2.9
13.3
10.1
1993
0.6
18.8
14.0
9.3
1994
0.6
18.4
14.0
8.5
1995
0.6
18.6
14.0
7.7
1996
0.6
18.5
14.0
7.0
1997
0.6
18.6
14.0
6.2
Alternative Vacancy Rate Paths
Optimistic8
No Crunch*
1988
9.8
9.1
1989
10.4
9.3
1990
11.1
9.7
1991
11.7
10.1
1992
12.3
12.5
1993
12.7
12.2
1994
13.2
12.0
1995
13.5
11.8
1996
13.8
11.5
1997
14.0
11.3
Notes: Base case uses actuals 1987-92. See appendix for further detail about these simulations.
a
b
Demand growth = 1.4% 1990-97, Construction = 29.0 1988-97
Demand growth = base case, Construction = base case + 4.0 Billion 1989-97
335
Causes and Consequences
Summary: Multi-family
The evidence suggests that the space overhang dominates any effect from the credit
crunch for the multi-family sector. Lending was not weak relative to construction activity during the alleged crunch period. All major lenders who had not already exited
the market cut back by similar amounts. Indicators of excess capacity, such as vacancy
rates, foreclosures, and prices, continued to erode throughout the alleged crunch period.
Although lending standards clearly tightened, it is hard to see that this had a substantive
impact on aggregate lending and aggregate construction activity. Even if capital was not
impaired (or equity investors were willing to recapitalize lenders) and loan rules were
not tightened, it is hard to imagine that construction would have taken a significantly
higher path, with a further accumulation of excess capacity.
There may have been some role for a crunch. There is little reliable data on local
vacancy rates so we cannot tell whether some markets were credit constrained. It is also
possible that in the absence of the crunch, the collapse in construction could have been
delayed a quarter or two and might have been a little less abrupt during the national recession. However, the sector is simply too small and the crunch effect had too secondary a role to argue for more than a few billion dollar crunch effect via this sector.
IV. Nonresidential Construction
Introduction
336
During the last three years, nonresidential construction has been the second weakest sector of the U.S. economy, declining over 25 percent from pre-recession peaks and showing little evidence of an upturn since the overall expansion began in 1991 (Chart 21).
As with the multi-family sector, nonresidential construction as a share of GDP reached
a record low in 1992 (1.5 percent). On the one hand, this underscores how deep the drop
in this sector has been; on the other hand, it shows that the sector is too small to have
more than a modest direct impact on the overall economy.
The decline in nonresidential construction has been by no means uniform across
types of structures. Construction actually increased in the industrial and institutional
(religious, educational, hospital and other) sectors. The collapse in activity has been
concentrated in three sectors; office buildings, hotels and motels, and wholesale and retail buildings more than accounted for the decline from 1986 to 1991. Together these
sectors comprise more than half of nonresidential construction.
Why did these three sectors collapse? As with the multi-family sector, we compare
the two principal explanations: overbuilding and the credit crunch. We examine financing trends, forecasts, model simulations and various measures of excess capacity in the
nonresidential construction industry. Due to data limitations, much of our analysis focuses on the office sector, but we also bring evidence to bear on the other weak performing sectors.
We find that overbuilding problems were so pernicious and pervasive that they dominate any other explanations for the construction collapse in these sectors. Financial
flow data do not indicate an unusual constraint on credit from depository institutions. A
variety of measures show a large and sustained build up in excess capacity in these sectors. Although the fall in the price of nonresidential buildings has been dramatic, it appears to be consistent with the fundamentals of weak demand. Cross sectional data
confirm virtually no metropolitan region had a need for additional office capacity. In
other words, virtually all nonresidential indicators suggest that the collapse in construc-
Chart 21: Nonresidential Construction Put-in-Place
$1987, billions
50
1972
74
78
80
82
84
86
88
90
92
Note: "Other" includes Industrial, Religious, Educational, Hospital and Institutional and Miscellaneous building classifications.
tion was a vain attempt to return the market to equilibrium and the credit crunch had, at
most, a small influence on the process.
Model simulations confirm this impressionistic evidence. Econometric models fail
to demonstrate an unusual shortfall in construction. Simulations of a simple accounting
model show that even in the absence of a recession, higher office building was not economically justified during the period of alleged credit crunch.
Given the persistently high vacancy rates of the period, the drop in lending during
the period seems to be a symptom, rather than a cause of the collapse in activity. We
conclude that the observed decrease in nonresidential construction was reasonable from
the perspective of profit maximizing individual investors, was consistent with normal
standards of prudential lending, and was desirable from the perspective of efficient resource allocation. In fact, even if nonresidential construction continues to remain depressed, it is likely that there will be an excess supply of space in the three overbuilt
sectors for much of the 1990s.
The Nonresidential Boom and Bust
As with the multi-family sector, any explanation of the collapse in nonresidential construction must start with the building boom of the 1980s. As Chart 21 shows, construction in all three key sectors—office, hotel/motel and retail/wholesale—reached record
heights in the mid-1980s.
In general, the same forces driving multi-family construction were behind the boom
in these sectors, but there were also some particular factors that added to the "froth," especially in the office market. Both the easing of depreciation rules under ERTA and the
liberalization of investment rules under Garn-St. Germain were particularly dramatic
for nonresidential real estate. Many academic and professional studies of real estate
during the 1980s focused on the office market and extrapolated low-risk high-returns
337
Causes and Consequences
based on data from a relatively short historical period.29 In addition, many builders and
lenders assumed that new communications technology and other changes in the demands on office buildings would spur demand for new and larger space, regardless of
the vacancy rates in older buildings.
Optimism was also fueled by a belief that demand for office space would continue to
grow rapidly and would be immune to recession. Even in the double-recession of the
early 1980s office employment grew, helping support a sharp increase in office construction. Indeed, as Chart 22 shows, in the typical cycle, office employment has grown rapidly up to the peak of the business cycle, leveled off during the recession and then
reaccelerated during the recovery. By contrast, in the current cycle office employment
growth was weak going into the recession, declined during the recession and has grown
very little during the recovery (even in comparison to the weakness elsewhere in the
jobs market).30
The collapse in the office market also had some unique elements. At the start of the
crunch, the flow of loans to the nonresidential sector was several times larger than loans
to the multi-family sector. Not surprising the losses from nonresidential loans and the
resulting damage to capital was larger as well. One estimate has write-offs by U.S.
banks for commercial real estate loan losses from 1989-91 amounting to $18 billion.
"Other Real Estate Owned" by commercial banks grew from just 14 billion in 1989 to
28 billion in 1991.3I Nonbank lenders suffered as well: by 1992 foreclosure rates for
commercial mortgages at life insurance companies had reached 3 1/2 percent, more than
double their mid-1970s peak.
The Case for a Credit Crunch
338
A strong case can be made that credit standards tightened in the nonresidential sector.
A variety of surveys and testimonials argue that individual projects to meet special
needs and in markets that are not overbuilt were starved of funds due to the credit
crunch. Starting in 1990 the Senior Loan Officer survey has included questions on real
estate investment. Although the responses have steadily moderated over time, it is well
known that a large proportion of the lenders tightened lending standards.
Credit crunch advocates also point to the sharp collapse in credit flows (Chart 23).
They first note that mortgage flows have fallen only modestly, but that this may be due
to financial institutions being forced to either refinance short-term construction and mortgage loans with new mortgages or face defaults by their borrower. They then point out
that construction loans, which are more reflective of the supply of funds to the market,
have contracted much more rapidly than construction activity. From the first quarter of
1990 to the middle of 1991 construction loans were slashed by over half, while construc29
For example, Browne and Case (1992) concluded that commercial construction, especially for office
space, is inherently cyclical and "the cycle of the 1980s was magnified by lax and institutional changes
and by a conviction—shared in by developers, banks, the academic community, and the general public that real estate was a high-return, low risk investment."
30
This optimism is understandable given the unpredictability of this sector. The consensus forecast as late
as the fall of 1991 was for a sharp reversal of the previous year's decline, followed by a gradual mending
process. It was only in 1992 that the prognosis shifted to a long convalesce, with little prospect of a complete recovery. The large and persistent errors by forecasters, using both models and judgment, suggests
that fundamentals do not adequately explain major swings in construction. Although the vacancy rate
looked dangerously high, there were few other warning signals— in terms of weak prices, illiquidity in
resales etc.—that a bust was around the corner.
31
Based on a survey by the American Banker.
Chart 22: Office of Employment at Business Cycle Peaks
Employment Levels
Index: NBER Peak=100
108
-24
-20
-16-12-8
-4
P
4
8
12
16
20
24
28
32
Months from Peak
Note: Average of 1960, 1969, 1973, and 1982 business cycles.
lion activity fell by less than a quarter. Putting these three pieces of circumstantial and
anecdotal evidence together—the complaints from borrowers, the pressures on lenders
and the contraction in credit—builds a plausible case for a binding credit constraint.
The Case for Overbading's Dominant Role
Overbuilding advocates see the other side of the coin in some of the crunch arguments.
They argue that like the multi-family sector, effective demand has fallen so dramatically
that in many markets only specialized new projects can yield an expected return sufficient to justify construction. In other words, in these markets the effective demand
curve has shifted below the minimum point of the supply curve, as in Case (3) in the
previous section. They argue that the anecdotal and survey evidence reflects not only
tightening of lending standards relative to the high-flying 1980s, but also the failure of
some borrowers to fully recognize that their creditworthiness has declined. The same
project that was creditworthy in 1985 may not be creditworthy in the current period of
less generous tax breaks, high vacancy rates and declining demand for space. They
point out that the sharp drop in credit flows is consistent with the idea that the cutback
in lending would be concentrated in the most speculative, highly leveraged projects.
Four kinds of evidence argue against a significant role for the credit crunch: financing patterns, indicators of excess capacity, cross-sectional vacancy rate data and model
simulations.
Financing Patterns
Under close examination, patterns in financing do not support the credit crunch story.
In particular, we find that depository institutions did not show signs of abnormal lending
constraint. All types of lenders, both regulated and nonregulated, cut back sharply on
339
Causes and Consequences
Chart 23: Nonresidential Real Estate Financing and Activity
$1987, billions
250
1
200
Mortgage Loans -§^
150
100
50
1970 72
74
76
78
80
82
84
86
88
90
92
their loans to the nonresidential real estate sector. It is possible that this reflects different
kinds of credit restraints that were equally binding for all major lenders. More likely,
however, it reflects a common factor—such as overbuilding or the recession—which affected the markets for all classes of lenders equally.
At first sight, the aggregate data appears consistent with an easing of credit followed
by a tightening such that the "normal" gap between construction activity and loans has
been reestablished (Chart 23). When the data are desegregated into depository and nondepository loans, however, the credit crunch hypothesis is rejected by the data. While
construction loans from depositories have tumbled dramatically, they have virtually disappeared for other lenders and the depository share is now near its all-time high (Chart
24). Similarly, depositories have seen their share of mortgage loans scale new heights
during the alleged crunch period. Of course, non-crunch factors may have offset the
crunch effects and these data are not very reliable;32 nonetheless, with every type of
non-depository (Life Insurance Companies, Mortgage Companies and Pension Funds)
cutting back on loans as much or more than every type of depository (Commercial
Banks, Savings and Loans and Mutual Savings Banks), one has to question this aspect
of the credit crunch hypothesis.
Excess Capacity
340
While the lending patterns across institutions are suggestive of a demand-induced drop
in construction, indicators of excess capacity are conclusive: the constriction in the supply of credit to nonresidential structures was rendered more or less irrelevant due to the
dramatic drop in demand for new space.
Chart 25 puts the high pace of office building in the mid 1980s into an historical perspective. Here, we graph the office stock per office employee and the downtown vacan-
32
For example, mortgage companies cut back in part because banks curtailed lending to them.
Chart 24: Depositories' Share of Nonresidential Loans
Percent
1970
72
92
Chart 25: Office Stock to Employment Ratio and the Office Vacancy Rate
$1987, thousands
28000
24000
10
20000
16000
12000
1970
72
90
92
Note: Vacancy data from Coldwell Banker. Office employment is calculated by applying
Torto/Weaton weights to major employment sectors. See text for details.
341
Causes and Consequences
cy rate.33 The pattern is clear. In the 1980s, office construction steadily outpaced office
employment growth, driving up the ratio of space to employment. The rising vacancy
rates show that most of the additional space was not being absorbed, yet developers continued to build. Even with the collapse in construction in the past two years, neither ratio
has declined to more reasonable levels.
The data available for the hotel/motel sector suggests a similar, and longer lasting,
excess capacity problem. As Chart 21 illustrates, hotel/motel construction was twice as
high in the 1980s as in the 1970s. In the face of this construction boom, the occupancy
rate trended down throughout the 1980s, both in the boom periods of 1983-84 and 198789, and in the periods of weakness. In 1991 the occupancy rate fell to 60 percent, the
lowest level in the history of this series, compared to the range of 62 to 66 percent in the
1980s.
Similar detail on other types of nonresidential space is not available, but the consensus view of industry analysts is that these markets are also suffering from severe excess
capacity problems. Chart 26 shows construction flows hit new highs as a share of GDP
in the mid and late 1980s for the office, retail/wholesale and hotel/motel sectors. Despite the subsequent decline in new construction, the stock of space continued to climb
for all three sectors. The chart also shows that the building stock as a share of GDP rose
through 1990, and then eased only slightly in 1991. Some of this rise undoubtedly reflects a desirable long-run process of capital deepening, and some may represent the
need for more space to accommodate new technology in offices and increased product
diversity at retailers. Much of the added space, however, appears to have little economic
justification.
Price Data
Although there is little reliable data on nonresidential real estate prices, anecdotal evidence and expert opinion paints a clear picture. A wide range of news reports suggest
that both the RTC and financial institutions have been selling properties at about a 50
percent discount.34 Of course, most of these sales are for relatively distressed properties.
In early-1992, Prudential Securities calculated the average decline in office building
prices to be 17 percent since 1991 and predicted further declines of 8 percent.35 Back
of the envelope calculations by Browne and Case (1992) suggest a greater drop in prices
is consistent with economic and tax fundamentals. Looking at a typical office building,
controlling for changing tax effects and using reasonable assumptions of rental and vacancy rates, they show that declines of over 50 percent from mid-1980s values are reasonable. This suggests that the observed price declines can easily be justified by
economic fundamentals. More important it confirms just how dramatic the drop in effective demand has been, and how plausible it is to assume that the demand curve now
lies below the break-even point for new construction.
Cross-sectional data
342
Until this point we have focused on data for the nation and broad regions. Of course,
real estate markets are inherently local in nature, segmented both geographically and by
33
Coldwcll Banker has collected data on suburban office vacancy rales only since the mid-1980s. Suburban
vacancy rates have hovered above 20 percent for their whole history.
34
See, for example, Barsky (1992), Klcege (1992), and Schmidt (1992).
35
Corcoran (1992).
type of structure. Therefore, although on average commercial real estate is clearly overbuilt, it is still possible that some markets are not severely overbuilt, leaving room for
localized credit crunch effects. With this in mind, Chart 27 presents a scatter plot for 44
Metropolitan Statistical Areas. Presumably for any given market if vacancy rates are
already low or are falling rapidly there may be room for additional construction. The
chart plots the 1988 vacancy rate against the 1988-91 change in the vacancy rate for
each area.36 As a frame of reference the unshaded area on the chart encompasses cities
with vacancy rates of 10 to 20 percent in 1991. The chart shows, for example, that the
36
1992 data was not available at the time of this analysis. Judging from the more up-to-date, but more
aggregated, national data adding another year would not change the conclusions.
Chart 26: Nonresidential Investment as a Share of GDP
Percent
10
Stock as a Share of GDP
I
1964
67
70
73
\VXVV
76
79
82
85
88
91
90
92
V
Construction as a Share of G w
1972
74
76
78
80
82
84
86
88
343
Causes and Consequences
Chart 27: Vacancy Rates by Metropolitan Area
Change in Vacancy Rate (1988-91)
15
Vacancy rate greater than 20% in 1991
10
Albuquerque
5
Oklahoma City
0
Sacramento
Honolulu
-5
Las Vegas
-10
Vacancy rate less than 10% in 1991
-15
0
5
10
15
20
Austin
25
40
1988 Vacancy Rate
Source: Coldwell Banker
vacancy rate for Hartford was about 8 percent in 1988 and rose about 10 percentage
points over the next three years. To put this in perspective, this compares to the average
national vacancy rate of about 10 percent in the 1970s.
As you would expect the scatter plot slopes down to the right: cities with high initial
vacancy rates tended to lower them and vice-versa. More importantly, the chart shows
just how few markets had room to build in the 1988-91 period. Only four cities—Sacramento, Columbus, Honolulu and Las Vegas—had reasonably low (below 11 percent)
and falling vacancy rates. Beyond these four cities, its hard to see where the credit
crunch could have constrained new office construction.
Model Simulation Results
As with residential construction, model simulations can give us a sense of whether a
"nonfundamental" factor such as a credit crunch was depressing nonresidential construction. We first explore formal econometric models and then turn to a simple vacancy
rate model.
Econometric Models
344
The first equation we looked at was the specification for total nonresidential structures
investment that is used in the MPS macroeconometric model. We then attempted to develop an improved equation that utilizes error-correction (ECM) concepts and apply it
to office construction only.37 The in-sample fit of the model was quite good, explaining
37
In the MPS model polynomial distributed lag (PDL) techniques relate past levels of business output and
capital costs to current levels of construction. The use of a PDL is based on a Jorgcnson investment function where there are lags between changes in fundamental conditions (the level of business output and capital costs in this case) and construction. The ECM framework explicitly incorporates long-run factors by
Chart 28: Nonresidential Structures Investment Revised MPS Model Results
$1987, billions
1970-86 Estimation Period
h 970-89
Estimation
Period
Dynamic Forecast using
1970-92 (full)
1970
72
84
86
88
90
92
about 98 percent of the variation in the ratio of investment to capital. Unfortunately, the
simulation results were quite sensitive to changes in the sample period. This problem
of poor out-of-sample results is common in nonresidential models and helps explain
why industry analysts did not provide early warnings to the overbuilding and high vacancy problems.38
Chart 28 shows predictions from the MPS equation when estimated over three different sample periods: the full sample (1970-1992), the pre-tax-reform period (19701986), and the pre-crunch period (1970-1989). 39 All three equations yield substantial
positive errors in the early and mid 1980s, confirming the overbuilding hypothesis. The
three simulations yield conflicting results for the current period: the 1970-86 version
predicts a sharp decline in construction, but not as deep as the actual collapse; on the
other hand, both other versions track the recent period quite closely. Although these results cast doubt on the credit crunch hypothesis, the generally poor fit over the 19781982 period and the sensitivity of the results to the sample period chosen point out the
danger of drawing strong conclusions from this modelling and forecasting exercise. 40
38
Footnote 37 continued
first fitting a trend line to the stock of buildings. A second equation models construction, using both the
effect of trend reversion forces that are based on the results of the first step (the error-correction adjustment) and additional short-run factors to explain cyclical fluctuations.
See Kopcke (1993) for evidence that conventional investment models fit structures construction substantially worse than they fit equipment spending.
39
Using data only from the partial sample is useful because it allows us to compare out-of-sample results to
in-samplc fits. All three full period forecasts were created using a dynamic process that fully simulates the
dependent variable. For this model, the dynamic forecasts do not include AR(2) effects that are included
in the estimated equations since calculation of the error terms requires knowledge of the actual dependent
variable.
40
More formally, diagnostic tests showed that both equations arc subject to instability problems.
345
Causes and Consequences
Results from ECM specification for office building construction also seem to confirm the overbuilding story. The top panel of Chart 29 shows the error term from two
different specifications. Thefirstsimply fits the stock of office buildings to a time trend.
The second fits the stock of office buildings-to-employment ratio to a time trend and the
MPS business cost of capital measure.41 Both equations attempt to capture the trend in
the stock of office buildings and yield similar patterns for the deviation of actuals from
trend. The bottom panel shows the results from using these trend deviation terms in
41
Because of multicollincarity, including bolh a time trend and (unrestricted) office employment as explanatory variables led to unstable parameter estimates.
Chart 29: Error Correction Model for Office Construction
346
Percent
20
Deviation of Stock from Long Run Trends
15
.Model 2
-
1
10
Model 1
is.V
5
0
-5
-10
J
\
-15
-20
72
74
76
78
80
i
i
82
84
86
88
90
92
Percent
3.5
Office Construction as a Percent of Existing Stock
Actual
3.0
Model 2
0.5
72
74
76
78
80
82
84
86
88
90
92
Note: Model 1 employs a long run equation between office stocks and a time trend. Model 2
employs a long run equation between stock per office employee, a time trend, and the MPS
cost of capital.
modelling short-run construction movements with an equation (along with some shortrun factors in an ECM framework that is consistent with the long-run trend assumptions).
Both long-run equations point out overbuilding in the 1980s, as seen by positive deviations between the actual stock and the trend level. The errors turn negative by 1992,
however, implying a shortage of office space. This conclusion is at odds with the high
vacancy rates discussed earlier and stems from a very strong time trend effect. Experimenting with different model specifications and trying various explanatory variables,
we found that a pure time trend effect dominates the office stock series and the results
are quite sensitive to whether breaks in this time trend is allowed. In particular, the surge
in construction in the mid-1980s—whether due to speculative excesses or a real need
for space—greatly affects estimates of long-run trends and therefore any conclusion
drawn from the model. As with the MPS model, the sensitivity of these results calls into
question any conclusion from this part of the model.42
Despite problems with interpreting the long-run equations, the short-run equations
track office construction reasonably well. In particular, both track the peaks and troughs
in the dependent variable quite well, explaining about 95 percent of the movement in
the dependent variable. The only additional explanatory variables, besides the errorcorrection term, are changes in business output, changes in capital costs, and the office
vacancy rate. The results in Chart 29 show that a decline in construction is consistent
with these fundamental forces, such that no additional "X factor" such as the credit
crunch is needed. However, because the estimated deviations from trend turned negative, an upturn in 1992 is predicted by both models. We believe these prediction errors
are due to overly optimistic trend lines, not credit-crunch effects.
Vacancy Rate Model
Given the estimation problems, it is difficult to draw conclusions from the formal econometric models. To get an idea of the magnitude of overbuilding and its implications for
new construction, we created a simple accounting model for downtown office space that
uses observed depreciation rates and the historical relationship between office space in
use and office employment growth to simulate vacancy rate paths under various economic assumptions for the construction and absorption of office space. (This model is
very similar to the vacancy rate model for multi-family housing discussed in Part III and
is also described in the appendix.) For out-year projections, the depreciation rate remains unchanged and office space per employee grows at a modest rate of 1.3 percent
per annum. Table 4 shows the results of three such simulations.
The table begins with our "baseline" scenario. In this case, the vacancy rate falls to
8.4 percent in 1997 after rising to the unprecedented level of 17.8 percent in 1992. This
simulation assumes no rebound in construction and the growth in office demand is equal
to population growth—in other words, no cyclical recovery in either demand or supply.
Looking back on the 1989-92 period, the baseline shows that the sharp drop in office
construction was swamped by an even greater decline in demand, resulting in a rising
vacancy rate. Looking forward, it shows that even at the current pace of construction it
will take four years to get the vacancy rate under 10 percent—a reasonable estimate of
the maximum long-run sustainable vacancy rate. This suggests that the actual plunge in
office construction has put the sector back on a path toward long-run equilibrium.
42
The sample period is not large enough to exclude the 1980s and then adequately estimate and run an outof-sample simulation.
347
Causes and
Consequences
Table 4: Office Sector Vacancy Model Simulations
348
Base Case and Realistic Forecast
Demand Growth
Absorption
Construction
Vacancy Rate
1977
4.7
11.6
11.6
0.0
1978
-3.3
2.4
13.8
6.2
1979
7.3
21.1
18.3
4.5
1980
8.2
24.2
23.4
3.7
1981
8.3
26.4
28.3
4.3
1982
6.7
25.0
35.3
8.0
11.6
1983
3.5
19.6
31.1
1984
6.6
28.0
37.5
13.9
1985
7.6
32.8
44.5
16.1
1986
6.8
33.2
38.8
16.6
1987
6.0
33.1
35.1
16.3
1988
5.3
32.7
36.0
16.3
1989
4.9
33.2
36.3
16.3
1990
2.8
27.4
32.1
16.8
1991
0.8
21.5
25.3
17.5
1992
-0.5
17.2
18.7
17.8
1993
2.1
26.6
18.7
16.0
1994
2.1
26.5
18.7
14.2
1995
2.0
26.6
18.7
12.4
1996
2.2
27.2
18.7
10.5
1997
2.3
27.7
18.7
8.4
Alternative Vacancy Rate Paths
Optimistic8
No Crunch13
1988
18.1
16.3
1989
19.2
16.3
1990
19.7
17.2
1991
19.9
17.8
1992
19.6
17.6
1993
18.9
17.3
1994
18.0
16.9
1995
16.8
16.4
1996
15.3
15.8
1997
13.5
15.2
Notes: Base case uses actuals 1987-92. See appendix for further detail about these simulations.
a
b
Demand growth = 5.3%, 1988-97; Construction = 44.5, 1988-97.
-Demand growth = 2.1%, 1991-97; Construction = 36.0, 1988; 36.3, 1989; 34.0, 1990;
30.7, 1991; 27.2, 1992-97.
The other scenarios in the table underscore the sheer unsustainability of the construction boom of the 1980s. For example, if we assume that the rapid demand growth of the
mid-1980s (which averaged 6.5 percent in the 1983-88 period) continued and that construction levels were kept at their 1985 peak, the vacancy rate would have ballooned to
19.9 percent in 1991 before trending down. Extending the employment boom forward,
the vacancy rate would not have fallen below 10 percent until the turn of the century.43
The second alternative assumes a constraining credit crunch that can be illustrated as
follows. First, suppose that the credit crunch accounted for half of the decline in construction from 1989 to 1992. Further, assume a middle-of-the road path for office employment growth after 1988 that is equal to the growth in the working age population.
Together this no-crunch, no-recession scenario results in a vacancy rate of 17.6 percent
in 1992, which is close to the actual rate of 17.8 percent. However, holding construction
constant from 1992 on, the vacancy rate would still be above 15 percent in 1997. Given
the current tax environment and reasonable investor expectations about the likely return
to investment, it is hard to imagine such a large, persistent and voluntary accumulation
of unused capacity.
Summary: Nonresidential Construction
In concluding this section it is important to be clear about the limits of the evidence we
have presented. First, although the loan data suggest no special constraint on depository
lending to this sector, this may reflect equally strong credit restraints for all lenders. Our
judgement, however, is that this argument is not compelling because many elements of
the credit crunch, including most of the legislative and regulatory actions of the period,
were specifically aimed at depository institutions. Second, while it appears that excess
capacity was so great that a collapse in construction was inevitable, the overbuilt sectors
had "lived with" this excess for some time, even after the shock of tax reform. It may
be, therefore, that the credit crunch played a key role in the timing of the collapse.
Again, it is our judgement that changes in market fundamentals—tax reform, the unexpected weakness in service jobs, the recognition of the severity of the excess capacity
and the likely low return to investment—were sufficient to trigger the collapse in construction. In the absence of credit constraints the collapse might have been delayed a
few quarters, but it might have been even more severe.
None of our arguments about the modest impact of the credit crunch on new construction contradicts the notion that real estate problems caused considerable collateral
damage—both literally and figuratively—to other sectors. All of the factors that drove
down real estate prices helped damage both the capital of lenders and the loan collateral
of borrowers. With few creditworthy new building projects available this "credit crumble" probably had little impact on the overall construction industry. We cannot deny that
it may have had a major impact on borrowers in more creditworthy sectors of the economy, however.44
43
The situation may be worse than the reported data suggest. Throughout the 1980s the "absorption" of
office space consistently outpaced the growth in office employment. In the first half of the 1980s this
upward drift probably reflected a desired trend to spend more on space per employee. After tax reform,
however, the excess space built probably began to reflect unwanted space that was still under rent (and
therefore "occupied") but not in actual use. If all of the growth in space per worker after 1986 is counted
as unused, the "true" vacancy rate in 1992 would be more than 25 percent!
44
Even here we need to be careful about how much "blame" we ascribe to the credit crunch. A large part of
the drop in real estate prices was due to the noncrunch factors we have been discussing. The resulting
"capital crumble" is the proximate cause of the drop in lending and economic activity; but the ultimate
349
Causes and Consequences
IV. Conclusion
350
In concluding it is useful to pare down the wide range of evidence and focus on a few
key findings. For the single-family sector we find that the weakness in construction cannot be easily explained by housing fundamentals. A close look at a range of indicators—such as the inventory of homes for sale and data on credit flows and credit terms
for different segments of the market—however, fails to uncover any sign of a credit
crunch. Since there are several other plausible explanations for the housing shortfall—
debt retrenchment, less sanguine home price inflation expectations and competition
from the glut of apartments—there is little reason to believe in a strong credit crunch
effect.
The argument for multi-family and nonresidential construction is more complicated.
In many respects these sectors were at the heart of the credit crunch: many of the legislative and regulatory changes were made in response to excesses in these sectors. Yet,
we find compelling reasons to believe the credit crunch did not have a binding impact
on new construction in many of these markets. Excess capacity was so widespread geographically, so resistant to declining construction and seems so likely to persist (given
the weak demand prospects) that it is hard to imagine a higher path for construction during the crunch period. To complete our crime investigation metaphor, the credit crunch
may have put a bullet into the heart of the real estate sector, but it had already been bludgeoned to death by a combination of excess capacity, adverse changes in tax law, weak
demand fundamentals and, because of all of the above, negative investor expectations.
The evidence does not rule out some limited credit crunch effects. The collapse in
real estate prices probably caused considerable collateral damage to other sectors of the
economy. In the single-family sector credit restraints on smaller builders may have
modestly depressed overall construction. In addition, the lack of a well-developed secondary market for "jumbo" loans may have caused a small increase in the interest rate
spread for those kinds of mortgages. In the multi-family and nonresidential sectors
credit restraints may have sped the adjustment to excess capacity, and there may have
been a few metropolitan areas or special markets where overbuilding was not a problem,
leaving room for binding credit restraints. The data simply are not rich enough to rule
out these possibilities.
As a final note, this paper's principle findings of a small role for the credit crunch and
a large role for excess capacity has important policy implications. In reviewing the policy options we start from the premise that the primary challenge is to ease the short-run
adjustment problems from excess capacity, without adding to that excess capacity. By
this criteria, an easing of tax and credit rules would be counterproductive. Tinkering
with the tax and regulatory code will not have a meaningful near-term impact on new
construction and any easing of the rules which would be sufficient to revive commercial
and multi-family construction would only divert more investment into economically unproductive uses. Even in the single family sector, where there are not large visible vacancy rates, it is hard to justify new tax or regulatory incentives. After all, this sector is
already heavily subsidized due to the deductibility of mortgage interest and various programs to support mortgage markets. In this light, the apparent efforts of households to
Footnote 44 continued
causes of the decline in activity are the factors that caused prices to fall in the first place. Adding another
layer of complication is the fact that the capital crumble is not the only channel from falling real estate
prices to declining lending. The fall in values also helped change investor and lender expectations about
the likely capital gains on new projects. This caused lower lending but should not be considered part of
the credit crunch.
reduce their real estate commitments seems like a positive move toward better resource
allocation.
There is another reason to argue against major policy actions: time is on our side.
As we have shown, it will take years before excess capacity is worked off in many markets and by that time credit crunch concerns should be far behind us. The mending process is already well under way in terms of bank capital. For example, among the 100
largest banks the average ratio of Tier I capital to risk adjusted assets rose from 8.5 percent in 1991 to 10.0 percent in 1992, matching the regulatory requirement for a "well
capitalized" bank. Our feeling is that any tendency toward over zealous regulation will
also fade over time. In other words, in terms of our supply and demand diagram, as the
demand curve shifts up the supply curve should recover as well, so that we may never
have a binding impact from the crunch.
Of course, the real estate contraction has caused considerable collateral damage. Is
this to be ignored as well? Here policy action is appropriate. In particular, all else equal,
it makes sense to ease overall macroeconomic policy to stimulate other sectors of the
economy and offset the impact of the real estate contraction. Stimulative macroeconomic policy will probably have little immediate impact on the overbuilt sectors, but it
will accelerate the long-run mending process. In the interim such a policy will help ensure that capital diverted from the over-built sectors ends up increasing other forms of
investment rather than lowering overall economic activity.
Appendix
Vacancy Rate Models
Both the Multi-family and Office Space models rely on the stock-flow identity:
S . M I - ^ S M + I,
(I)
For example, in the Multi-family model S is the stock of apartments, 8 is the net rate at
which old structures are abandoned and I is new construction. The growth in the demand, or absorption, of apartments is a function of weighted population growth: p,
which accounts for apartment use rates for each age group and a drift factor: a, which
captures the change in per-person space use:
D . s D ^ O + p . + o,) (2)
The vacancy rate is then simply:
V, = (S,-D t )/S l
(3)
To make the model operational, we use historical series for V, P, and I and an estimate
of 5 to calculate the implied history of D, S, and a. Then by assuming different future
values for a and I plus projections of p derived from census projections, we calculate
the future path of V for these various scenarios.
For the nonresidential or office space model, we use an estimated series for the number of office workers as our primary determinant of demand (p). These estimates use
are based on a TortoAVheaton study that calculated the shares of the major employment
sectors to "office" employment: 6.7% of manufacturing; 15.0% of mining; 3.0% of construction; 6.0% of wholesale and retail trade; 1.5% of transportation, communication
and public utilities; 25.2% of services; and 100% of FIRE (finance, insurance and real
estate) industries.
Additional insight into these models can be gained by substituting (1) and (2) in (3)
and rearranging to get:
Vt=l-(l-VM)[(l+pt
351
Causes and Consequences
and:
V,= l - ( l - V l . 1 ) | l + p l + a l - g l |
where g, = I/S, - 5 is the net investment rate. Here, we see that changes in the vacancy
rate are due to a net factor that combines demand and supply growth. Therefore, the
scenarios that we report are really dependent on assumptions about the combination of
p, a, and g and not individual components. In other words, simultaneously raising p,
one percent, lowering (X, one-half a percent, and raising g, one-half a percent would not
change the V, projection. (Further detail on these simulations are available upon request
from the authors.)
References
352
Akhtar, M. A., and E. S. Harris (1986-7). Monetary Policy Influence On The EconomyAn Empirical Analysis. Federal Reserve Bank of New York Quarterly Review. Winter 1986-87, pp. 19-34.
Barsky, Neil (1992). Commercial Real Estate Sales Perk Up. Wall Street Journal.
March 11, 1992, p.A2
Bernanke, Ben S., and Cara S. Lown (1991). The Credit Crunch. Bmokings Papers On
Economic Activity. No.2, pp. 205-48.
Boldin, Michael D. (1993). Econometric Analysis of The Recent Downturn in Housing
Construction: Was it a Credit-Crunch? Federal Reserve Bank of New York Research
Paper no. 9331.
Browne, Lynn E., and Karl E. Case (1992). How The Commercial Real Estate Boom
Undid the Banks. Real Estate and the Credit Crunch. September 1992.
Browne, Lynn E., and Eric Rosengren, eds. (1992). Real Estate and the Credit Crunch.
Conference Series No. 36, Boston: Federal Reserve Bank of Boston, 1992.
Campbell, Michele, and Denise Polmar (1991). Evidence of the Credit Crunch. Housing Economics National Association of Homer Builders, March 1991.
Carlton,Jim (1992). Real Estate Developers Feel Pressure of Tight Lending, Wall
Street Journal. January 17, 1992, p. A7c.
Corcoran, Patrick (1992). How Deep is the Bottom in the Commercial Office Market?
Economic Review (The Prudential Insurance Company of America, April 1992).
Corker, Robert, Owen Evans, and Lloyd Kenward (1989). Tax Policy and Business Investment in the United States: Evidence from the 1980's, International Monetary
Fund Staff Papers. Vol. 36, Number 1, March 1992, pp. 31-62.
DiPaquale, Denise, and Jean L. Cummings (1992). Financing Multi-Family Rental
Housing: the Changing Role of Lenders and Investors Housing Policy Debate Vol.
3, Issue 1, 1992, p. 77-81.
Esaki, Howard, and Judy Wachenheim (1985). Explaining The Recent Level of Singlefamily Housing Starts. Federal Reserve Bank of New York Quarterly Review. Winter 1984-85, pp. 31-38.
Greenhouse, Steven (1992). Fed Chief Links Tight Credit to Slowdown, New York
Times. November 19, 1992, p. Dl.
Gyourko, Joseph, and Richard Voith (1993). Leasing as a Lottery: Implications for Rational Building Surges and Increasing Vacancies, Journal of the American Real Estate and Urban Economics Association. Vol. 21, No. I, Spring 1993, pp. 83-106.
Hendershott, Patric H. (1991a). Are Real Housing Prices Likely To Decline By 47 Percent? Regional Science and Urban Economics. Vol.21, pp. 553-63.
Hendershott, Patric H. (1991 b). An Altered U.S. Housing Finance System: Implications
for Housing NBER Working paper no. 3770, July 1991.
Hendershott, Patric H., and Edward J. Kane (1992a). Case and Consequences of The
1980S Commercial Construction Boom. Journal of Applied Corporate Finance.
Vol. 15, No.l (Spring), pp. 61-70.
Hendershott, Patric H., and Edward J. Kane (1992b). Office Market Values During Past
Decade: How Distorted Have Appraisals Been? Working Paper Series College of
Business, The Ohio State University. Working Paper 92-68.
Hester, Donald D. Financial Institutions and the Collapse of Real Estate Markets, in
Browne, Lynn E. and Eric Rosengren eds. (1992) Real Estate and the Credit Crunch.
Conference Series No. 36, Boston: Federal Reserve Bank of Boston, 1992.
Kleege, Stephen (1992). S&L's Gain Popularity with Developers, Survey Finds American BankerMarch 3, 1992, p. 11.
Kopcke, Richard W. (1993). Forecasting Investment with Models and Surveys of Capital Spending New England Economic Review, Federal Reserve Bank of Boston.
March/April 1993, pp. 47-69.
Litan, Robert E. Banks and Real Estate: Regulating The Unholy Alliance, in Browne,
Lynn E. and Eric Rosengren eds. (1992) Real Estate and the Credit Crunch. Conference Series No. 36, Boston: Federal Reserve Bank of Boston, 1992.
Mankiw, N. Gregory, and David N. Weil (1988). The Baby Boom, The Baby Bust, and
the Housing Market. Regional Science and Urban Economic. Vol. 19, pp. 235-58.
Mankiw, N. Gregory, and David N. Weil (1991). The Baby Boom, 'Pent-up1 Demand,
and Future House Prices. Journal of Housing Economics. Vol. 1, pp. 347-67.
Mosser, Patricia C. (1992). Changes in Monetary Policy Effectiveness: Evidence from
Large Macroeconometric Models Federal Reserve Bank of New York Quarterly Review. Spring 1992, pp. 36-51.
Peek, Joe, and Eric S. Rosengren (1992). Crunching The Recovery: Bank Capital and
the Role of Bank Credit. Real Estate and the Credit Crunch. September 1992.
Polmar, Denise (1991). AD&C Financing Survey Results, Housing Economics National Association of Home Builders, June 1991, p. 10-14.
Poterba, James M. (1991). House Price Dynamics: The Role of Tax Policy and Demography. Brookings Papers On Economic Activity. No.2, pp. 143-83.
353
Causes and Consequences
354
Poterba, James M. (1992). Tax Reform and the Housing Market In The Late 1980s:
Who Knew What, and When Did They Know It? Real Estate and the Credit Crunch.
September 1992.
Ryding, John (1989). Housing Finance and the Transmission Mechanism of Monetary
Policy. Studies On Financial Changes and the Transmission of Monetary Policy
New York: Federal Reserve Bank of New York and Federal Reserve Bank of New
York Quarterly Review (Spring 1990).
Schmidt, Susan (1992). RTC Sales Said Well Below Book Value Washington Post. September 9, 1992, p. Fl.
Seiders, David (1991). The Outlook Housing Economics National Association of
Homer Builders, January 1991. p. 1-4.
Summers, Lawrence (1987). Corporate Capital Budgeting Practices and the Effects On
Tax Policies On Investments. In Martin Feldstein, Ed., The Effects of Taxation On
Capital Accumulation. Chicago: University of Chicago Press.
Throop, Adrian (1986). Financial Deregulation, Interest Rates, and the Housing Cycle.
Federal Reserve Bank of San Francisco Economic Review. Summer 1986, pp. 63-78.
Wojnilower, Albert M. (1980). The Central Role of Credit Crunches In Recent Financial History. Brookings Papers On Economic Activity No.2, pp. 277-326.
Wojnilower, Albert M. (1985). Private Credit Demand, Supply, and Crunches—How
Different Are the 1980's? The American Economic Review. May 1980, pp. 351-56.
Credit Supply Constraints on Business Activity,
Excluding Construction
by Charles Steindel and David Brauer1
This paper explores possible effects of the pronounced slowdown since 1989 in credit
growth, particularly from banks, on nonfinancial business activity excluding construction. Businesses routinely use various financial intermediaries and credit markets in order to finance both current operations and long-term projects.2 How individual firms are
affected by a disruption in the supply of bank credit, which may arise either from a radical change in the terms of funding or from outright credit rationing, will depend both
on the firm's economic environment and on its access to alternative funding sources.
Firms facing such a disruption and unable to obtain credit from other sources could
be forced to cut costs in order to conserve working capital. We may therefore observe
cutbacks in any of a number of activities, including overall production and employment,
capital spending, inventory accumulation, and expenditures on research and development. The fact that in 1990 and 1991 both bank lending and employment fell in most
industries does not, however, establish a direct or significant causal linkage between the
two. With the economy in recession, a decline in activity is to be expected, and the decline in economic activity may itself limit firms' demand for credit.3
One potentially fruitful approach to the credit constraint issue is to examine small
and large firms separately. A growing body of academic literature focuses on the hypothesis that the connection between small firm credit problems and their activity can
explain significant elements of cyclical downturns—a hypothesis one might expect to
hold with special force during the recent episode. Small firms tend to rely on bank credit
to a much greater extent than do large firms, and thus may be especially vulnerable to a
bank credit crunch (Radecki 1990). One reason for their vulnerability is their limited
1
Our thanks to M.A. Akhtar, Michael Boldin and Ethan Harris for comments. Joseph Abate and Stuart
Sclatcr-Booth provided extensive assistance with this paper.
2
Strategic uses of finance—"financial engineering" as a profit center, and financial restructuring in the
course of reorganization—will not be discussed in this paper.
3
Other papers in this study examine the relationships between the credit slowdown and the economic environment.
355
Causes and Consequences
access to close substitutes for bank credit (Gertler and Gilchrist 1993). Access to commercial paper and other securities markets tends to be limited to large firms. A credit
crunch would likely have its greatest impact on firms with a high rate of investment demand relative to cash flow (Fazzari, Hubbard, and Petersen 1988). Such firms are likely
to be small and fast-growing.
Thus, we can examine various measures of small firm activity and see how the recent
experience compares with earlier cyclical downturns, keeping in mind the relative mildness of the current episode. An alternative approach is to examine the behavior of firms
or industries that ended the 1980s with relatively strong or weak financial positions. We
might hypothesize that even if all firms have equal access to various sources of credit,
those with a high ratio of debt to equity or a high interest burden could react more
strongly to a loss of any single funding source than those in a stronger financial position.4 Thus, we can check whether the relative contraction of activity in already distressed firms was worse in recent years than in earlier downturns both with and without
adjustments for the different macroeconomic environment.
The first part of the paper documents movements in borrowing and output in several
highly aggregated business sectors. All the sectors saw unusually sharp retrenchments
in borrowing during the early 1990s but output measures tended to be in line with past
periods around recessions, suggesting that the weakness in borrowing was in great part
due to supply problems.
As a preliminary step toward gauging any impact of credit restrictions, we next examine survey evidence. The early 1990s saw a moderate increase in the frequency of
complaints about credit tightness from smaller firms—especially when the overall state
of the economy is taken into account—and rather more extensive evidence that banks
tightened lending standards.
The balance of the paper uses a variety of sources to explore relative movements in
activity across firms scaled by size and measures of financial conditions. We find that
neither the absolute nor the relative performance of smaller firms was notably poorer
over the last few years than during past periods of overall economic weakness. Across
a number of dimensions, however, firms in weak financial shape did do worse in recent
years than in the past. Thus, our overall conclusion is that credit constraints may well
have played a significant role in impeding business activity in recent years. We find little
or no evidence to suggest that these constraints were an unusually severe burden for
small business, although we emphasize that all of our evidence in this area is indirect.
I. The Slowdown in Lending and Business Activity
356
This section looks, impressionistically, at trends in lending and activity in highly aggregated sectors. The purpose is to identify any unusual weakness in activity in recent years
that appears to be correlated with weakness in borrowing. Any such correlation might
suggest that credit constraints inhibited both borrowing and activity. Three sectors are
examined: All nonfarm nonfinancial business, nonfarm nonfinancial corporations, and
nonfarm nonfinancial noncorporate business.
Other papers in this study discuss the overall slowdown in business borrowing in
more detail. The key point for this paper is that movements in cyclically sensitive components of business activity, such as corporate cash flow and spending on inventories
4
Cantor (1990) found that capital spending by firms with relatively high ratios of debt to assets showed
greater sensitivity to cash flow changes.
and fixed capital can account for only a portion of the slowdown in business borrowing.
Because a very substantial fraction of the slowdown in debt growth remains unexplained by these basic nonfinancial demand factors, the possibility exists that the supply
of credit to business was cut significantly (see Mosser and Steindel, in this volume).
Charts 1-3 examine activity and debt measures for the three sectors we are considering.5 Each chart contains three panels. Panel (A) compares a sector's activity, measured
as real output, over the most recent business cycle with the average of the four preceding
cycles. The data are illustrated such that the reading at the business cycle peak equals
100. Data are presented from six quarters before the business cycle peak to ten quarters
after. The most recent peak occurred in July 1990, but because real GDP fell in the third
quarter of 1990, we treat the second quarter as the peak. Thus, the data from the current
cycle cover the period from 1988-IV to 1992-IV. Panel (B) presents a similar comparison for the real value of debt owed by the sector. Panel (C) charts the historic values
of the debt-to-output ratio for the sector.
The first sector examined is aggregate nonfinancial business.6 Chart 1A shows that
5
Sec Braucr (1993) for a more extensive discussion of economic activity for the period since 1989.
6
Aggregate nonfinancial output is the sum of real nonfinancial corporate output plus nonfinancial noncorporate output. In turn, noncorporate output is measured as the difference between nonfarm nonhousing output and corporate output, divided by the deflator for nonfarm nonhousing output, less real personal
Chart 1: Total Nonfarm Nonfinancial Business
Index
112
Index
1 10
A: Output
Average of Four
Previous Cycles
B: Debt
Average of Four
Previous Cycles
*^
110
105
•
—
100 -
y
pi
^ ^ ^ ^
95
Current Cycle _
on
-5 -4 -3
1959
-2 -1 P
61
63
1 2 3 4 5 6
Quarters from Peak
65
67
69
7
8
9 10 11 12
71
73
-5 -4 -3 -2 -1
75
77
79
81
P
1 2 3 4 5 6
Quarters from Peak
83
85
87
7
89
8
9
10 11
91
93
357
Causes and Consequences
the pattern of output in this sector since the 1990 peak has been in line with that of past
cycles, although growth going into the peak was weak. Chart I shows the data on the
cyclical performance of this sector's debt.7 The dramatic shrinkage of real debt in this
cycle, in contrast to its sustained growth in earlier cycles, is clearly evident. Chart 1
plots the debt to output ratio for the sector. The continued high level of the ratio is clear,
and although the ratio has shown a pronounced decline its fall does not stand out as
greatly against history; the decline in the ratio in the mid-1970s was comparable.
In the corporate sector output since the last peak has generally been on a par with the
average of past cycles (Chart 2A), although growth was unusually slow going into the
peak. Real corporate debt has fallen since the last cyclical peak, but the shortfall has
been less severe than for business as a whole (Chart 2B). Finally, the recent decline in
the corporate debt-to-output ratio seems in line with earlier periods (Chart 2C). It does
not appear that in the corporate sector as a whole—which is, of course, dominated by
Footnote 6 continued
consumption expenditures on professional medical and legal services. The last adjustment is made
because movements in these services, while a component of noncorporate output, are clearly governed by
very different forces than businesses such as retailing. The noncorporate and total nonfinancial output
series also encompasses the output of noncorporate financial business, including real estate firms.
7
The debt includes that owed by medical and legal partnerships and proprietorships, (see note 6, above).
Chart 2: Nonfarm Nonfinancial Corporate Business
Index
Index
no
1 IU
A: Output
Average of Four
Previous Cycles^
110
B: Debt
108
Average of Four
Previous CycleSy^
106
104
102
105
100
98
100
Ni^
J%^^
Current Cycle
^^^
96
94
.<y
1
OR
-5 -4 -3 -2 -1 P
1952
358
1 2 3 4 5 6
Quarters from Peak
7
8
9 10 11 12
Current Cycle _
1
1
1
1
-5 -4 -3 -2 -1
•
P
i
i
i
i
1 2 3 4 5 6
Quarters from Peak
•
•
•
•
7
8
9
88
•
i
10 11
92
large firms—that debt problems are associated with weakness in activity.8
The noncorporate sector, dominated by smaller businesses, tells a very different story. Chart 3A shows that noncorporate output has tended to track above its average path
in this cycle. However, by late 1992 the shortfall of real debt widened to the neighborhood of 30 percentage points (Chart 3B), and the decline in the ratio of debt to output in
this sector was unprecedented (Chart 3C). It is unlikely, then, that demand factors can
by themselves account for the weakness in debt growth in this sector. This disparity suggests that disruptions in the supply of credit probably played a significant role in the
weakness in this sector's debt growth, and this may have worked to hold down activity.
The impressionistic evidence of this sector supports the notion that smaller firms
may have borne a disproportionate share of the retrenchment in debt. It does not necessarily follow from this, though, that a "credit crunch" centered on smaller firms weakened activity in this sector, given the relative strength of noncorporate activity. To get
a firmer handle on small business problems, the next section of the paper examines survey evidence on credit conditions for smaller firms.
1
The output and debt performance of manufacturing—which is dominated by corporations- -has been
roughly similar to that of nonfinancial corporations as a whole.
Chart 3: Nonfarm Nonfinancial Noncorporate Business
Index
130
B: Debt
Average of Four
Previous Cycles
120
110
100
90
-5 -4 -3 -2 -1
P
1 2 3 4 5 6
Quarters from Peak
7
8
80
9 10 11 12
-5 -4 -3 -2 -1
P
1 2 3 4 5 6
Quarters from Peak
7
8
9
10 11
250
200
150
100
1959
61
63
65
67
69
71
73
75
77
79
81
83
85
87
89
91
93
359
Causes and Consequences
Survey Evidence on Lending to Smaller Firms9
360
The most comprehensive survey covering smaller firms is conducted by the National
Federation of Independent Business (NFIB), whose over 5(X),(XX) member firms together account for roughly half of total private sector output and employment. Most members are small to medium-sized and consequently can be presumed to have limited
access to nonbank capital markets.
The NFIB survey includes questions about activity measures such as employment,
price changes, inventories, capital expenditures, and the cost and availability of credit.
With regard to credit, members are asked whether they normally borrow at least once
every three months. Those who are regular borrowers are asked what interest rate they
paid on their most recent loan, how that rate compared with the rate paid three months
earlier, and whether such loans were "easier or harder to get than they were three months
ago." l() In addition, the survey asks "What is the single most important problem facing
your business today?," and lists "interest rates & financing" as one of nine answers."With this survey, then, it should be possible to detect periods in which credit
availability as perceived by borrowers either worsened or improved significantly.12 We
identify periods during which many small businesses complained about credit availability and cost.
Despite the anecdotal evidence of a credit crunch, results of recent NFIB surveys offer surprisingly little direct evidence that businesses have perceived unusual difficulty
in obtaining credit during the last several years. Between mid-1990 and mid-1992 the
net percent of respondents reporting that credit was harder to get (percent harder minus
percent easier) ranged from 11 to 14 percent (Chart 4). These figures, however, were
lower than in previous recessions and much lower than in 1980. Perceived credit conditions have been unusually stringent in New England since early 1990 and in the Southwest since 1986, but in other regions credit availability problems have been less
frequently reported than in previous recessions. Breakdowns by industry show unusual
financing difficulties only in the wholesale trade sector, and even then only to a limited
degree relative to past recessions. By early 1993 the net percent reporting that credit
was harder to get had fallen to just 8 percent.
At the same time, the available credit became less costly over the 1989-92 period.
Short-term interest rates paid by businesses declined from an average of 11.7 percent in
July 1989 to 5.0 percent in December 1992, and less than 5 percent of respondents have
reported paying higher rates at any time since 1990. In contrast, rates were much higher
and rising during the 1979-82 period, when nearly all regular borrowers reported higher
rates during several quarters (Chart 5).
Finally, in the last few years relatively few firms have considered interest rates or
9
10
1
A more exhaustive analysis of such evidence can be (bund in the paper by Hamdani, Rodrigucs. and
Varvatsoulis in this volume.
Note that we cannot directly compare perceived current credit conditions with those of earlier years, only
the extent to which they have recently worsened.
' Emphasis in original. The other choices offered are "taxes," "inflation.11 "poor sales," "cost of labor,"
"government regulation(s) & red tape," "competition from large businesses," "quality of labor," and
"shortage of fuels, materials or goods." Respondents also have the option of writing in a different answer
to this question.
12
Respondents tend to accentuate the negative. Since the survey's inauguration in 1973 the number of members reporting easier credit conditions has never exceeded the number reporting that credit was harder to
obtain.
Chart 4: Small Business Credit Problems
Percent Harder To Get - Percent Easier
Net Percent Reporting "Harder"
40
1973
75
Chart 5: Interest Rates on Commercial and Industrial Loans
Percent
50
Percent
40
Actual Rate
(right scale)
30
20
10
^
Percent Paying Higher Rate
I. I I
-10
1973
75
77
79
81
83
. (lefts.cale) . I
85
87
89
91
-5
92
Sources: Quarterly NFIB Survey and FRB Release E.2 "Survey of Terms of Bank Lending."
Notes: Actual rate is the effective average interest rate on short-term commercial and industrial loans. Shaded areas represent recessions.
credit availability a serious problem. In the fourth quarter of 1992 only 3 percent listed
interest rates and financing as the most serious problem facing their business, and that
361
Causes and Consequences
figure has not exceeded 7 percent since 1989. These results can be compared with figures as high as 37 percent during the 1982 recession. As a group, small businesses appear to be much more concerned with taxes, regulation, and weak demand. The NFIB
itself has concluded that the survey findings for all areas except New England cast doubt
on the existence of a credit crunch (Dunkelberg and Dennis 1992). The organization
points to the decline in the fraction of respondents seeking to borrow as evidence that
the main problem has been lack of demand.
Nevertheless, the survey results should be judged in light of the overall economic situation. The 1990-91 recession was much less severe than that of 1981 -82. In the earlier
period, rapidly plunging sales would have cut many firms off from credit without any
change in supply conditions. An important criterion in judging the existence of a credit
crunch is the continuing availability of credit to firms in comparable financial shape.
This suggests that we should adjust the credit survey responses to the general economic
environment. Hamdani, Rodrigues, and Varvatsoulis (this volume) find that such an adjustment boosts the recent credit availability readings considerably. In fact, the adjusted
reading in 1989 was much higher than in the 1973-75 recession and not far short of the
levels seen in 1982. This adjusted reading suggests that there was a fairly significant
curtailment of credit to many, mostly small, businesses.13 Appendix 1 summarizes some
of the recent structural impediments to small business lending.
At first glance, the survey evidence from the lender's point of view offers little solid
evidence that the contraction in the supply of bank credit was more pronounced for
small than for large business. The responses from the Federal Reserve's periodic survey
of Senior Loan Officers show a considerable tightening of the terms of lending for all
businesses in 1990 and 1991, but little difference between small and large firms in the
degree of tightening (Board of Governors 1992). However, because small firms are likely to have fewer alternatives to bank loans than do large firms, what appears to be an
equal degree of tightening from a bank's viewpoint could ultimately lead to a more pronounced degree of contraction in overall small business credit and activity.
All in all, then, the surveys provide some evidence of a significant tightening in credit conditiQns to smaller firms, which may well have held down activity. In the next section of the paper we focus on activity in smaller firms to see if any weakness is evident
in the disaggregated data.
Evidence on Size Effects
362
Evidence that smaller firms have underperformed larger firms in recent years would be
suggestive, at least, that credit crunch effects have been important, given the evidence
of significant tightening of credit to smaller firms. We will look at four sets of data: employment in small- business-dominated industries; production, employment, and capital
spending, trends in manufacturing industries ranked By size; Quarterly Financial Report
data on lending to manufacturing firms ranked by size; and Compustat data on employment, research and development (R&D) spending, and inventory accumulation in firms
ranked by size.
13
Regression analysis of ihc relationship between nonfarm noncorporate output and the Hamdani, Rodrigucs, and Varvatsoulis adjustment to the credit availability index found a modest negative relationship
between the adjustment and output but this disappeared when the relationship was corrected for seriallycorrelated errors.
Small Business Employment
As a first pass at evaluating small firm performance, we look at employment trends in
the two and three digit industries that the Small Business Administration (SBA) has
identified as being small- business-dominated. The SBA defines a small business industry as one in which a minimum of 60 percent of employment is in firms with fewer than
500 employees, and a large business industry as one in which 60 percent of employment
is in firms with 500 or more workers. Although we cannot produce consistent time series that precisely replicate the SBA aggregates, we can make reasonable approximations.
According to our series, employment in the small business sector fell 1.2 percent between July 1990 and March 1991, while other private employment declined by 1.6 percent (Table 1). We find that small-business-dominated industries experienced more
modest employment losses in 1990-91 than during the three earlier recessions, but that
their performance relative to other industries was moderately weaker than in the 197375 and 1981-82 recessions. Small business industries did exhibit unusually weak employment growth during the first two years of the recent recovery. This weakness could
reflect firms' difficulty in obtaining the credit necessary to finance expansion. Nonetheless, other industries also experienced much weaker than normal employment growth
during the recovery.
Small Establishment Manufacturing Industry Performance
Next we examine data for the manufacturing sector by two digit industry.14 Although
this breakdown is somewhat cruder than the one based on the SBA definition, it offers
14
The term "two-digit" is a bit of a misnomer, since we can decompose the transportation equipment industry into its two three-digit components: motor vehicles and other transportation equipment.
Table 1: Employment Growth
Small-Business-dominated and Other Industries
Small-Businessdominated Industries8
Other Private Industries
A. Recession
November 1973-March 1975
-2.6%
-4.0%
January-July 1980
-2.0
-1.5
July 1981-November 1982
-2.1
-4.2
July 1990-March 1991
-1.2
-1.6
March 1975-March 1977
8.7
6.0
November 1982-November 1984
11.7
7.1
March 1991-March 1993
1.4
0.1
B. Early expansion
Source: Bureau of Labor Statistics, seasonally adjusted by authors.
a
Based on SBA definition (see text).
363
Causes and Consequences
much richer data on activity. We focus on manufacturing as the largest sector that exhibits a strong cyclical pattern. Table 2 ranks the two-digit manufacturing industries
by their average establishment size. 15 We look at how four measures of industry activity—industrial production, payroll employment, real investment, and the real inventory to sales ratio, have behaved relative to establishment size over three recent periods:
1980-83, 1984-88, and 1989-91. 16 These periods encompass two recessions and the
bulk of the long 1980s expansion. We want to know whether activity has contracted
15
The industrial production, inventory-sales ratio, and payroll employment data arc the standard Bureau of
Economic Analysis (BEA) series used at a quarterly frequency. The investment data is the annual BEA
series on constant-dollar spending on nonresidential fixed capital used in compiling the data on capital
stock by industry. (The data on plant and equipment spending, which are available quarterly, have limited
two-digit coverage).
16
Because the average establishment size in manufacturing industries has been little changed we use the
1972 size as the standard for all periods.
Table 2: Mean Establishment Size in Manufacturing Industries, 1972
Employees per Establishment
364
Lumber
20
Printing
25
Miscellaneous manufacturing
29
Stone, clay, and glass
39
Nonelectric machinery
45
Furniture
50
Fabricated metals
51
Food
56
Apparel
56
Rubber
67
Petroleum
69
Chemicals
73
Instruments
76
Leather
85
Paper
105
Textiles
132
Electrical machinery
135
Primary metals
168
Other transportation equipment
169
Transportation
195
Motor vehicles
238
Tobacco
244
more in the recent period for small firm industries than for large firm industries, and
whether the relative small firm industry performance has been worse recently than in
the past.
Results are shown in Charts 6-9. The vertical axis of each chart measures the cumulative percentage change in activity over each period, while the horizontal line measures
average establishment size. Each point plotted represents an industry. The information
we are trying to extract from the charts relates to the relative performance of small and
large industries in each period, as opposed to their absolute performance. In other
words, wish to know whether growth for an activity in a period tended to be greater for
large firms than for small firms. We are not concerned here with whether or not the cumulative growth of the activity variable was positive for the industry during the period
in question.
A simple way to summarize the relative performance of small and large firms is to
calculate the regression line fitting the points in each scatter diagram. If the regression
line slopes up, it suggests that large industries tended to show greater cumulative growth
for an activity during the period shown. A negative slope suggests better performance
for smaller firms. The solid lines in Charts 6-9 are these regression lines. To give some
notion of the statistical reliability of the slope of the regression lines, we include in the
charts dashed lines with slopes greater and less than the regression line by a magnitude
of one standard deviation from the estimated slope.17 Chart 6 summarizes the evidence
for industrial production as it relates to establishment size. The negative slope of the
regression line in each period indicates that on the whole industries with the fastest
growth (or smallest decline) in output tended to be smaller than average, although in no
case is the negative relationship statistically significant. This relationship appears most
clearly in the recent period, suggesting that small establishment industries did not experience unusual weakness relative to large establishment industries. Chart 7 shows that
small establishment industries likewise did not reduce employment by a greater amount
than large establishment industries, either in the most recent period or in the past. Charts
8 and 9 show positive but insignificant connections between establishment size and both
investment and inventory behavior in all periods.
To this point, the activity measures suggest little or no relative contraction of small
establishment industries in the 1989-92 period. The above analysis, however, did not
normalize for external demand conditions facing the industry. As a useful, albeit imperfect, proxy for demand for the industry's products we use the performance of industrial production by industry.18 We derive predicted levels of employment and capital
spending by industry by regressing them against industry output over the 1980-92 pe-
17
The regression calculations provided both the slope and the standard error of the slope estimate. We then
added this standard error to the slope estimate and drew a line with this new slope through the point representing the means of the horizontal and vertical values of the points plotted through the scatter.
The new line is plotted through the point of means for convenience. A regression line, by construction, must go through this point, and this procedure gives some visualization of the possible ranges of
regression lines summarizing the data.
18
Industrial production has some (laws as a demand measure. An industry can in the short run use inventory
swings to break the connection between demand and production, and in the long run, production and
demand can drift apart because production is an index of value added (sales less purchased inputs), and the
fraction of industry sales representing value added at the manufacturing level can change. Nonetheless,
production is probably a very good proxy for demand when we look at horizons of a year or two, which is
the focus of this paper. Data on real shipments, an alternative demand measure, are not available for many
industries.
365
Causes and Consequences
Chart 6: Percentage Change in industrial Production against Average
Establishment Size
% Change in IP
1980-83 Period
Motor Vehicles
20
10
-
— * 7! - - - -•
0
^
•
10
—
-
#
Tobacco
20
• Primary Metals
Petroleum
on
0
100
50
150
200
250
300
% Change in IP
80
1984-88 Period
60
-
• Machinery
Other Transportation
40
#
20
-"»
c--j
•
1.
0
-20 h
-
Leather
-40
150
100
0
50
% Change in IP
15
200
250
300
1989-92 Period
10
5 ;^
*
-
t
0
"•
Tobacco
-5
#
-10
Motor Vehicles
-15
-20 #
-25
366
50
**--*,
-
Leather
100
150
200
Avg. Establishment Size
250
300
Chart 7: Percentage Change in Employment against Average Establishment Size
% Change Employment
30
1980-83 Period
20
Fabricated Metals
10
Tobacco
0
-10
-20
-30
-40
Primary Metals
0
50
100
200
150
250
300
% Change Employment
30
1984-88 Period
20
Other Transportation
10
0
-10
Tobacco
-20
-30
-40
Leather
0
100
50
150
200
250
300
% Change Employment
o
1989-92 Period
Food •
*•
•
r\
Paper
U
Tobacco
-5
10
•
"""*•••.
Electrical
i
1C
50
100
150
200
Avg. Establishment Size
250
300
367
Causes and Consequences
Chart 8: Percentage Change in Investment against Average Establishment Size
368
% Change Investment
1980-83 Period
Tobacco
120
100
Miscellaneous
-
80
60
40
r- - • - *: v
0
•
20
y
•
-20
0
50
% Change Investment
100
-'
"
•
—
:
Motor Vehicles
#
i
i
i
i
100
150
200
250
300
1984-88 Period
Tobacco
80
Furniture
60
40
20
Motor Vehicles
0
-20
• Food
-40
0
50
% Change Investment
100
150
200
250
300
1989-91 Period
30
-
Furniture
Tobacco
20
10
t
"t
0
• Textiles
10
•
• Food
90
50
100
150
200
Avg. Establishment Size
250
300
Chart 9: Change in Inventory/Sales Ratio against Average Establishment Size
Change in I/S
1.0
1980-83 Period
0.8
Other Transportation
0.6
0.4
0.2 -
•
Petroleum
0.0
-0.2
.
«
-0.4
-0.6
Motor Vehicles
• Rubber
1
0
50
100
•
•
i
150
200
250
200
250
Change in I/S
0.2
1984-88 Period
0.0
• Paper
----•-._
-0.2
-0.4 b- - * "
-0.6 Primary Metals
-0.8
-1.0
• Machinery
0
50
100
Change in I/S
0.6
1989-92 Period
0.4
150
Primary Metals
• Paper
0.2
0.0
-0.2
-0.4
Machinery
Other Transportation
-0.6
-0.8
50
100
150
Avg. Establishment Size
200
250
369
Causes and Consequences
370
riod. I9 The period of estimation is intentionally kept short to avoid consideration of
long-term structural changes in factor demand. For inventories we follow a slightly
different procedure. Because ratios of inventories to sales in most of the manufacturing
sector have exhibited a pronounced downward trend in recent years, we derive a normalized predicted ratio of inventories to sales from time trends.
These regressions give benchmarks for industry factor demands. To relate the shortfall or excess of actual activity from these benchmarks to average firm size by industry,
we plot the cumulative residuals for these regressions over the different periods against
the industry's average employment per establishment. For any industry the (cumulated)
error for the entire 1980-92 period will be zero (since the average difference between a
regression prediction and an actual value will be zero, the (cumulated) difference, or error, will also be zero). In any subperiod, however, a regression can under- or overpredict
actual values. We hypothesize that if credit restrictions are important then unusual
weakness in activity will be more likely to occur in small establishment industries, especially during recessionary periods.20
Charts 10-12 show the scatters. As in Charts 6-9 we show an estimated line summarizing the data, as well as lines with slopes differing from the estimated line's by one
standard deviation. These estimated lines are derived from cross-sectional regressions
summarizing the data shown in the charts; they are separate from the time-series regressions for each industry that are used to produce the points in the data scatter.
Chart 10 plots the cumulative residual change in employment over each of the three
periods against average establishment size. A positive reading for any industry in any
period would imply that employment growth was unusually strong in that industry given
actual industrial production during that period—in other words, a positive reading indicates that the regression relating employment growth to industrial production underpredicted employment growth in the period. As in Chart 7, we find that during the 198083 period a positive employment "surprise" was more likely to occur in industries dominated by small establishments—the solid line shown in the chart slopes down. This
negative relationship disappeared during the mid-1980s but reappeared more strongly
than before during the most recent period. In other words, large-establishment industries experienced greater employment retrenchment relative to production in 1989-92
than did small-establishment industries.21 Charts 11 and 12 repeat this exercise for capital spending and ratios of inventories to sales. Capital spending shows a weak positive
relationship between establishment size and activity over the last few years, while the
relationship between establishment size and surprises in the ratios of inventories to sales
was strongly negative. Thus, when demand is controlled for, there is no strong evidence
that industries dominated by small firms saw unusual contractions in activity during the
most recent period. Like Charts 6-9, Charts 10-12 show little sign that the slopes of the
lines summarizing the data are significantly different from zero.
19
The employment equations were estimated in growth rates; the capital spending equations were estimated
in log-levels. In estimation we occasionally obtained a negative coefficient on current industrial production. Where this was the case, the equation was rcestimatcd using lagged industrial production. If this
coefficient was also negative the industry was excluded from the analysis. The regression coefficients and
summary statistics arc tabulated in Appendix 2.
20
It is possible that recent restructuring by large corporations, which has involved contractions in their factor
demands, has tended to reduce any positive relationship between the residuals and size. One response to
this objection is that the regressions are estimated over the entire sample period, so the predicted levels of
activity already at least partly incorporate the influence of restructuring.
21
Alternatively, productivity growth was particularly strong in large-establishment industries.
Chart 10: Sum of Errors of the Percentage Change in Employment against the Pecentage
Change in Industrial Production Plotted against Average Establishment Size
Employment Error
20
1980-83 Period
Primary Metals
15
Motor Vehicles
Tobacco
.10
0
100
50
150
200
Employment Error
15
1984-88 Period
250
300
Tobacco
10 •
5 -
Leather
0
-5
-10
-15
0
50
Employment Error
10
1989-92 Period
"
-*
VjL^i;
^ --""*"•
## # •
"
i
Motor Vehicles
100
200
250
300
150
200
Avg. Establishment Size
250
300
150
• Electrical
-5
-10
50
100
371
Causes and Consequences
Chart 11: Sum of Errors of the Log of Investment against the Log of Industrial
Production Plotted against Average Establishment Size
372
Investment Error
0.5
1980-83 Period
Motor Vehicles
-0.5
-1.0
Petroleum
-1.5
-2.0
• Miscellaneous
0
50
100
200
150
250
Investment Error
1.5
1984-88 Period
• Miscellaneous
1.0
Motor Vehicles
-0.5
50
100
Investment Error
0.8
1989-92 Period
• Petroleum
0
150
200
250
0.6 h
0.4 -
Motor Vehicles
0.0
-0.2
. - - - "V"
•
•
""
~-
•»
i
-0.4
#
—
0.2
0
50
Apparel
100
150
Avg. Establishment Size
200
250
Chart 12: Sum of Errors of Inventory/Sales against Time Plotted against Average
Establishment Size
I/S Error
1.0
1980-83 Period
Other Transportation
•
0.5
\
0.0
"
.
•
-0.5 h
—
-1.0
_
Primary Metals
-1.5
Rubber
-2.0
-
0
50
Motor Vehicles
100
200
150
250
I/S Error
1984-88 Period
Motor Vehicles
4
Rubber
Primarv Metals
3
2
1
0
-1
Other Transportation
-2
50
0
100
I/S Error
1.5
1989-92 Period
1.0
•
Electrical
•
•
0.5
i
200
150
250
•
Other Transportation
0.0
-0.5
-1.0
•
•
-1.5
Rubber
•
-2.0 -2.5
50
Primary Metals
100
150
Avg. Establishment Size
** ^ ^
200
250
373
Causes and Consequences
Small Manufacturing Firms: Borrowing and Activity
We next look at data from the Commerce Department's Quarterly Financial Report for
Manufacturing, Mining and Trade Corporations (QFR). The QFR data provide income
and balance sheet information, and are broken down by firm size for the manufacturing
sector as a whole. For most manufacturing industries this information is also supplied
at the two-digit level. Moreover, the QFR data enable us to identify industries with unusually high exposure to bank debt. These industries, including textiles, paper, fabricated metals, and petroleum, should be the most vulnerable to tightening bank lending
standards. Compared with the data discussed in the previous section, the QFR data allow us to compare developments among small and large firms, as opposed to small- and
large-establishment industries.
The aggregate data show that bank lending to manufacturers did contract sharply
over the last several years, but not to a significantly greater extent than in past recessions
or early stages of recovery (Table 3). During the 1990 recession (with the second quar-
Table 3: Commercial and Industrial Loans Outstanding, Manufacturing Firms, by Source
Annual Average Percent Change
374
1987 Dollars 8
Current Dollars
Bank
Nonbank
Total
Bank
Nonbank
Total
1990-11 to 1991-1
3.0
6.4
5.3
-1.5
1.7
0.7
1981-111 to 1982-1V
3.1
4.0
3.7
-2.1
-1.2
-1.5
1980-1 to 1980-111
5.8
20.9
16.1
-3.6
10.3
5.8
1974-1 to 1975-1
14.1
16.6
15.7
3.0
5.3
4.4
Average of
previous three
7.7
13.8
11.8
-0.9
4.8
2.9
1991-1 to 1992-111
-3.5
-1.2
-1.9
-6.0
-3.8
-4.5
1982-1V to 1983-1V
-4.4
-1.8
-2.6
-8.0
-5.6
-6.3
1980-111 to 1981-111
21.0
13.5
15.7
10.1
3.2
5.2
1975-1 to 1976-1V
-10.0
8.6
2.3
-15.5
2.0
-4.0
2.2
6.8
5.1
-4.5
-0.1
-1.7
1983-1V to 1990-11
14.6
12.4
13.1
10.4
8.3
8.9
1976-1V to 1980-1
16.9
12.1
13.5
8.0
3.6
4.9
814.0
213.0
458.6
671.6
Recessions
Recoveries
Average of
previous three
Expansions
Memo: Loans outstanding in billions of dollars
1992-111
a
258.1
555.8
° Based on deflation using implicit GDP deflator.
ter of 1990 again treated as the cyclical peak), nominal bank lending rose at a modest
3.0 percent annual rate, a sharp slowdown from its growth rate during the 1980s expansion. In real terms, outstanding bank loanvS fell at a 1.5 percent annual rate, a slightly
smaller decline than during the 1980 and 1981-82 recessions. During the first six quarters of recovery, the rate of decline in bank lending accelerated. This pattern is consistent with the experience following the 1973-75 and 1981-82 recessions. During the
1990-91 recession the slowdown in bank lending was partially offset by a small increase
in real credit from nonbank sources. From the trough through late 1992 nonbank lending also declined, though not as much as in the aftermath of the 1981-82 recession.
Nonetheless, even though more than two-thirds of manufacturers1 total outstanding
credit is derived from nonbank sources, taking such lending into consideration does not
markedly alter our characterization of the latest period as one of unusually weak credit
growth.
In examining lending by firm size, we define small and medium firms as those with
under $250 million in assets in 1990. We follow Gertler and Gilchrist (1992) in retabulating the QFR data so that at all times large firms represent the same proportion of total
manufacturing sales as in 1990.
Chart 13 shows real bank lending to both size classes has declined during the last
several years, but similar declines also occurred during earlier recessions. Large firms
experienced a sharper decline in borrowing from banks than did small firms. A closer
look at the long-term pattern of small firm finance reveals a slow upward trend in bank
lending, but a decline in nonbank lending in the late 1970s (Chart 14A). In recent years,
a moderate contraction in bank credit beginning in 1989 interrupted a sustained uptrend
in small firm debt that began in the expansion of the mid-1980s. Small firm debt in 1991
was about $11 billion (roughly 8 percent) below this trend. Large firm debt was also
below its mid-1980s trend, but the shortfall was not as great in percentage terms (Chart
14B).
Chart 13: Bank Debt by Firm Size
150 -. 1
1
i
100 -
I
1
i
JT-
yyyyyyy
I
200 -
Total - / ^ t
y/yy/yy
Billions of 1987 dollars
*
y%
J
1
Large
/ 1
Small
1
50 -
H
1959
63
67
71
75
• 1
79
I
83
|
-
1.
87
91
93
375
Causes and Consequences
The observed reduction in loan volume, especially relative to the earlier trend for
small businesses, is consistent with the existence of a credit crunch. However, it does
not constitute proof of one or establish that any crunch affected real activity. In particular, the QFR data do not enable one to distinguish between a change in banks' ability
and willingness to finance business activity, and a shift in businesses' demand for bank
credit. The only direct measure of activity provided by the QFR involves inventory
stocks. Although the data indicate that small manufacturing firms do indeed hold fewer
inventories relative to sales than larger firms, this pattern has not changed markedly over
the latest decade.
Chart 14: Type of Debt by Firm Size
376
Billions of 1987 dollars
160
140
120
100
80
60
40
20
^$H_
1974
M.
76
78
80
82
84
86
88
90 91
Billions of 1987 dollars
700
1983-89 Trend Line
Bank Debt
1974
76
80
82
84
86
Source: Quarterly Financial Report and Gertler and Gilchrist (1992).
Note: Shaded areas represent recessions.
88
90 91
Data from Compustat
In a final set of exercises, we analyze firm-level data from Standard and Poor's Compustat data base to see if patterns of employment or other measures of real activity can
be explained by firm size. In using the Compustat data we would ideally like to develop
a sample consisting of a wide cross section of firms, offering a rich variety of data, and
spanning a number of years. Unfortunately, firms enter and drop out of the data base,
and the quality of many of the reported variables is poor. As a compromise, we selected
sets of data from three time periods spanning recessions: 1972-75, 1980-83, and 198891. A firm in the manufacturing and trade sectors was chosen for the sample in each period if it had plausible data for the following variables: employment, sales, and assets.22
For each of the three periods we begin by stratifying the sample by firm size during
the initial year of the period, and then examine movements in several variables of interest over the subsequent three-year period. Results of this exercise offer little evidence
that activity in the smaller firms has been unusually weak during the recent period. In
fact, average employment in the lowest quartile grew a surprising 48.5 percent between
1988 and 1991, far outpacing the growth in real sales. 23 By contrast, employment
growth was considerably weaker than sales growth in the other quartiles (Table 4). 24
22
Note that we do not include firms that may have dropped out of the sample because of bankruptcy. However, if the assets and operations of the bankrupted firms were acquired by other firms in the sample, they
are implicitly taken into account.
23
Because of late reporting, the 1991 sample is considerably smaller than that of the previous three years.
This disparity, however, does not affect measures of average employment relative to sales, since both measures are based on the same set of firms in each .year.
24
Results in these tables are based on stratification by net sales in the initial year. With the exception of
inventories, alternative stratifications by either assets or employment in the initial year do not significantly
alter the results.
Table 4: Employment and Sales Growth, by Firm Size in Initial Year
Selected Three-Year Periods, Percent Change
Quartiles8
Low
Second
Third
High
1972-75
14.4
14.1
7.6
1.3
1979-82
13.7
-1.4
-3.7
-7.7
1988-91
48.5
1.4
-3.7
-5.6
1972-75
27.8
22.5
20.3
21.7
1979-82
19.0
-0.5
-1.3
-8.2
1988-91
20.1
17.7
0.9
0.5
Average employment
Average real salesb
Source: Compustat.
a
Based on sales in initial year.
- Deflated by GDP deflator.
b
377
Causes and Consequences
During the earlier periods, employment growth tended to lag behind sales, with no particular pattern by firm size. Even during the recession, and with average sales flat, average employment in the lowest quartile grew by 22 percent between 1990 and 1991. In
that year employment and sales essentially moved in tandem in the other quartiles, as
was the case for all quartiles during previous recessions (Table 5).
Other activity variables offer only weak evidence that smaller firms have fared unusually poorly during the most recent period. In all quartiles, reported R&D spending
grew in real terms, but the most rapid increase took place in the two middle quartiles.
In the past there has been some tendency for R&D expenditures to grow fastest in small
firms (Table 6). Since 1988, ratios of inventories to sales have fallen in all size quartiles
based on initial year sales, but especially in the lowest one. However, the relationship
between size and the decline in the inventory-to-sales ratio disappears under alternative
size stratifications. Regardless of the stratification, no such pattern appeared during either of the earlier periods.
Summary of Small Firm Effects
The data in this section provide little support for the proposition that small firms have
experienced unusually steep declines in activity in recent years. Employment in industries dominated by small firms fell less during the 1990-91 recession than in previous
recessions, although the proportion of the decline accounted for by these industries was
somewhat greater than in the 1973-75 and 1981-82 (but not 1980) recessions. The twodigit data, the QFR data, and the Compustat data show no striking difference between
small and large firm performance.
IV. Financial Distress Effects
In this section, our approach generally parallels that of the previous section, but here we
Table 5: Employment and Sales Growth, by Firm Size in Initial Year
Recession Years, Percent Change
378
Quartiles3
Low
Second
Third
High
1974-75
-1.1
-2.5
-3.1
-4.1
1981-82
-6.9
-0.4
-5.2
-7.8
1990-91
21.9
0.3
-4.2
-7.6
1974-75
•4.5
-4.2
-3.5
-6.2
1981-82
•8.0
•2.8
-6.2
-6.4
1990-91
-0.6
2.2
-3.4
-9.1
Average employment
Average real sales*
Source: Compustat.
a
Based on sales in initial year (1972, 1979,1988).
- Deflated by GDP deflator.
b
focus on indicators of financial strength. If a credit crunch was a significant factor impeding activity in recent years, it is plausible to expect that businesses in a weak financial position suffered disproportionately. We will look at two of the data sets used in the
last section: the two-digit data on manufacturing activity and the Compustat data on
corporations.
Two-Digit Manufacturing Data
Manufacturing industries with very high ratios of net interest to cash flow or output may
plausibly be viewed as distressed. Thus, we can compare and contrast the relative performance of distressed and nondistressed industries with evidence from earlier recessions. Given the starting conditions for any industry, we want to know whether activity
has contracted more in high-net-interest industries than in low-net-interest industries.
We examine how the same four measures of industry activity used earlier—industrial production, payroll employment, real investment, and the real inventory-sales ratio
have behaved relative to the ratio of net interest to cash flow in the 1980-83, 1984-88,
and 1989-91 periods. Results are shown in Charts 15-18, which essentially repeat
Charts 6-9 but use the start-of-period ratio of net interest to cash flow on the horizontal
axis, rather than establishment size. 25
Chart 15 summarizes the evidence for industrial production as it relates to net interest. The downward slope of the solid line in the top panel of Chart 15 indicates that on
the whole, industries with strong output growth during 1980-83 did in fact start the period with lower than average ratios of net interest to cash flow. However, the corresponding line for the 1989-92 period slopes upward. Similarly, Chart 16 shows that past
negative relationships between employment and the ratio of net interest to cash flow
were not evident in 1989-92. Charts 17 and 18 show little or no sign of significant negative relationships between investment or the ratio of inventories to sales and the ratio
of net interest to cash flow.
These results provide little evidence that financially distressed manufacturing industries have contracted activity to a greater extent than others in recent years. Again, however, we need to control for external demand conditions facing the industry, using
industrial production as the primary measure of demand. Thus, we repeat last section's
25
All data are from the BEA. Cash flow is defined as the sum of pretax profits, capital consumption and proprietors' income.
Table 6: Average Research and Development Spending, by Firm Size
Percent Change
Quartiles8
Low
Second
Third
High
1972-75
62.7
35.8
39.5
39.5
1979-82
110.2
69.3
67.2
51.7
1988-91
29.8
90.1
25.5
29.3
Source: Compustat.
a
Based on initial year sales.
379
Causes and Consequences
Chart 15: Percentage Change in Industrial Production against Net Interest/Cash Flow
380
%Change IP
1980-83 Period
Machinery
Motor Vehicles
Primary Metals
-30
•0.15
-0.10
-0.05
0.15
0.00
0.05
0.10
1979 Net Interest/Cash Flow
0.20
0.25
%Change IP
80
1984-88 Period
60
• Machinery
40
J
Motor Vehicles
20
•
•
0
•
•
Primary Metals
;
-20
#
-40
-0.2
i
0.0
i
0.2
%Change IP
i
0.4
0.6
0.8
1983 Net Interest/Cash Flow
1989-92 Period
10
1.0
1.2
1.4
Rubber
i
5 •
0
-
'
-
^
:
-
-5
Motor Vehicles
•
10
•
-
15
20
• Leather
0*
-0.6
-0.4
-0.2
0.0
0.2
1988 Net Interest/Cash Flow
0.4
0.6
Chart 16: Percentage Change in Employment against Net Interest/Cash Flow
%Change Employment
• Fabricated Metals
1980-83 Period
10
*•
-10
_ Motor Vehicles
-20 -
-
-30 -
Primary metals
-An
-0.15
-0.10
-0.05
0.00
0.05
0.10
1979 Net Interest/Cash Flow
0.15
0.20
0.25
%Change Employment
30
1984-88 Period
Printing
i ^ ^ - ,
Primary metals
Leather
-40
-0.2
0.0
%Change Employment
1988-92 Period
-20
-0.6
-0.4
0.2
0.4
0.6
0.8
1983 Net Interest/Cash Flow
1.2
1.0
1.4
Food
-0.2
0.0
0.2
1988 Net Interest/Cash Flow
0.4
0.6
381
Causes and Consequences
Chart 17: Percentage Change in Investment against Net Interest/Cash Flow
382
%Change Investment
1980-83 Period
Tobacco
100 -
•
Miscellaneous
50 -
Other Transportation
0
•
• *
'
, ^ - *"
.en
50
i
-0.15
-0.05
-0.10
i
0.00
0.05
0.10
1979 Net Interest/Cash Flow
i
0.15
0.20
0.25
%Change Investment
100
1984-88 Period
• Tobacco
80
• Furniture
60
•
•*
40
*
^
^
^
•
20
Primary Metals
•
••
-20
-40
-0.2
0.0
-
0.4
0.6
0.8
1983 Net Interest/Cash Flow
0.2
1.0
1.2
1.4
%Change Investment
30
1984-88 Period
20
• Furniture
- *^^
:
10 •
•
* ^^
—r~
^
Petroleum
•
Motor Vehicles
-10 •-
Food
-20
-0.6
-0.4
•
-0.2
0.0
0.2
1988 Net Interest/Cash Flow
0.4
0.6
Chart 18: Change in Inventory/Sales against Net Interest/Cash Flow
%Change in I/S
I.U
1980-83 Period
Other Transportation
0.8
0.6 -
-
0.4 -
Petroleum
0.2 -
^
nn
••
••
^
U.U
'
0.2
0.4
^
••>
• •
•
•"
:
—
Rubber
. - - * " " " " • Motor Vehicles
•
n«
-0.15
-0.05
-0.10
0.00
0.05
1979 Net Interest/Cash Flow
0.10
0.20
0.15
%Change in I/S
0.2
1984-88 Period
0.0
Apparel
-0.2
-0.4 h
T
r
r
:
:
-0.6
i
i
0.0
0.2
%Change in I/S
0.6
1989-91 Period
^
—
* ^^
—
•
Primary Metals
Machinery
•
-0.8
^
* - - .
i
0.4
0.6
0.8
1983 Net Interest/Cash Flow
1.2
1.0
1.4
Primary Metals
0.4
0.2
0.0
-0.2
-0.4
• Machinery
Other Transportation
-0.6
-0.8
-0.6
-0.4
•
-0.2
0.0
0.2
1988 Net Interest/Cash Flow
0.4
0.6
383
Causes and Consequences
exercise, as summarized in Charts 10-12, only here we plot a period's shortfall or excess
in activity relative to that predicted by industrial production against the industry's initial
net interest burden, rather than establishment size.
Chart 19 summarizes the evidence for the employment data against net interest. For
each of the three periods we plot the cumulated residual employment change—over and
above that explained by industrial production growth—against the initial ratio of net interest to cash flow. Recall that a positive reading for any industry implies that employment growth was unusually strong given the actual performance of production. We see
little sign that industries with above-average ratios of net interest to cash flow in 1979
exhibited weaker than expected employment growth in 1980-83, and in the mid-1980s,
a positive relationship appears. The most recent period saw a weak negative relationship, albeit one with little formal statistical significance. In other words, high net interest industries saw greater employment retrenchment relative to production over the last
few years.26 Interestingly, petroleum—an industry that the QFR data records as unusually dependent on bank debt—showed an unusually large decline in employment relative to levels predicted by that industry's production, even though in an absolute sense
it was one of the stronger industries in terms of employment growth.
We repeated this exercise for capital spending and ratios of inventory to sales.27 Both
of these variables appeared to be positively related to the ratio of net interest to cash flow
during the current period and were not negatively related in the past. Thus, when demand is controlled for, evidence that industries facing high interest burdens have shown
unusually weak activity during the last few years appears to exist only for employment.
Compustat Data
Because industry statistics may mask important differences between firms within an industry, we once again turn to Compustat for firm-level data, using the same three periods
as in the previous section. To highlight firms that may have been unusually constrained
by debt and thus particularly vulnerable to tightening credit conditions, we now stratify
firms on the basis of the debt to asset ratio in the initial year. This ratio itself fell significantly among the most debt-ridden firms during the most recent period, but there is little
evidence that this has translated into weaker activity. Between 1988 and 1991, employment growth trailed sales growth in the quartile with the highest debt-to-asset ratio, but
by less than in the two lowest quartiles (Table 7). The top quartile was in fact best able
to maintain both employment and sales growth in the face of recession in 1991. By contrast, the top quartile suffered the greatest employment losses (both absolute and relative
to sales) during the 1982 recession. There was no sign of a relationship between debt
and either inventories or R&D spending in this or past recessions. Thus, firms particularly vulnerable to credit tightening on the basis of initial indebtedness do not appear to
have been unusually affected by a "credit crunch."
Summary of Financial Distress Effects
384
This section produced mixed results on the relationship between measures of financial
distress and business activity. The two-digit industry data suggest that in recent years
the more debt-burdened industries contracted employment by an unusual amount, rela-
26
An alternative view is that their productivity growth was surprisingly strong.
27
The relevant charts are available from the authors.
Chart 19: Sum of Errors of the Percentage Change in Employment against the Pecentage
Change in Industrial Production Plotted against Net Interest/Cash Flow
Employment Error
1980-83 Period
Primary Metals
15
10
Petroleum
•
— ^ *
M
-^
Tobacco
-10
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0.25
1979 Net Interest/Cash Flow
Employment Error
15
1984-88 Period
• Tobacco
10
Petroleum
/ •
-15
-0.2
0.0
0.2
0.4
0.6
0.8
1983 Net Interest/Cash Flow
1.0
1.2
1.4
Employment Error
10
1988-92 Period
-
Tobacco
•
-
••
•
-5
Petroleum
-10
-0.6
-0.4
-0.2
0.0
0.2
1988 Net Interest/Cash Flow
0.4
0.6
385
Causes and Consequences
tive to their contraction in production. Other measures of activity show little or no sign
of unusually large contractions by heavily indebted firms or industries.
V. Quantifying Size and Distress Effects Using the Compustat Data
The two sections preceding produced only modest evidence that either small firms or
firms with large debt burdens experienced unusually large contractions in activity during the recent recession. In this section we use the Compustat data to carry out somewhat more formal tests of the relative effects of size and financial stress effects.
As an alternative to the sorting by size and debt-to-asset ratio, we can perform regression analysis to formally test whether size and debt-to-asset ratios are important determinants of a firm's activity. We can also see whether the effects of these variables
have shifted over time. Accordingly, for the three time periods 1972-75, 1980-83, and
1988-91 we estimated simple regression models of employment, inventories, capital
spending, and R&D spending.28 These three periods encompass the last three prolonged
recessions and the initial stages of recovery (actually, with 1980 included, the last four
28
Cantor (1990) and Boldin (1992) used the Compustat data to test more general hypotheses about the determinants of capital spending by firm and their relationship to financial variables. To that end, they estimated more fully elaborated models and used panel data over substantial periods. Their models can be
used to address such questions as: Given well-established trends, which firms would have been expected
to have weak capital spending in 1989-91, and did they? For our purposes, this analysis would need to be
extended to address the further issue of classifying the weak firms by financial distress or size, and comparing that breakdown to earlier recessions. We chose to address these latter issues through direct reduced
form equations relating activity to sales and measures of initial firm size and financial distress. This
approach also sidesteps the vexatious issue of the degree to which a firm can by itself change such variables as its cash (low (say through renegotiation of labor contracts and refinancing of debt) in response to
an unfavorable financial environment, such behavior clouds the use of Cantor and Boldin's structural models in the evaluation of historic episodes.
Table 7: Employment and Sales Growth, by Debt Exposure
Selected Three-Year Periods, Percent Change
386
Quartiles*
Low
Second
Third
High
1972-75
7.0
3.5
8.9
-0.0
1979-82
6.4
8.7
1.3
-10.7
1988-91
14.1
-5.7
12.1
15.5
1972-75
30.8
20.2
29.1
16.6
1979-82
2.7
13.6
2.2
-6.0
1988-91
43.0
9.8
6.8
23.2
Average employment
Average real salesb
Source: Compustat.
a
- Based on initial year debt-to-asset ratio.
- Deflated using GDP deflator.
b
recessions of any length), and comparisons can show how the relationship between firm
activity and firm characteristics during downturns has changed. We also included estimates for the 1984-87 time period to see whether the relationships significantly differ
between expansions and contractions.
Our model proposes that a firm's activity is determined by both employment and
debt-to-asset ratios in the initial year. If initial leverage has a significant impact on decisions, such that heavily indebted firms curtail their activity more than other firms during recessions, we would expect this coefficient to be negative. Similarly, if, other
things being equal, smaller firms tend to shrink more in recessions than larger firms, we
would expect a positive coefficient. If leverage or size effects have been more important
recently, then the coefficient in the 1988-91 period should be of greater absolute magnitude than in earlier periods. The regressions also include year and industry dummy variables as proxies for aggregate shocks and industry-specific factors. The model is very
similar in spirit to the cross sectional regressions on the two-digit industry data, although the details differ because of differences in the type and quality of the data available.
Results for employment are shown in Table 8. They suggest that smaller firms suffer
more extensive employment losses than larger firms during periods around recessions.
In particular, the coefficients imply that a firm that enters a recessionary period with a I
percent smaller work force than an otherwise similar firm will experience job losses
over the period about .2 percent a year greater. In contrast, the relationship between employment growth and initial debt-asset ratios \s positive, albeit statistically insignificant,
in all three recession periods. There is no sign of a marked change in these patterns in
the more recent period. On the whole, these results suggest that smaller firms, other
things equal, have suffered a greater loss of employment than have large firms over the
last few years. However, there seems to be no evidence that more indebted firms have
contracted jobs to a greater extent recently than have less indebted firms.
Table 8: Determinants of Employment in Compustat Sample
Initial Employment
Initial Debt-toAsset Ratio
R2
1. 1972-75
.197
(.019)
.330
(.266)
.078
2. 1979-83
.241
(.019)
.048
(.286)
.104
3. 1988-91
.287
(.025)
.115
(.324)
.127
4. 1984-87
.205
(.022)
-.254
(.330)
.084
Period
Model:
Log Ejjt - Log E ijM = a log Eij0 + b(D/A)ij0 + cj + dt + e.
Ejjt = employment in firm i of industry j in year t.
D/A = debt to asset ratio.
Cj = industry dummy (= 1 if firm is in 2-digit industry j; 0 otherwise).
dt = dummy for year t.
e = constant term.
Note: Standard errors in parentheses.
387
Causes and Consequences
In the expansion period of the mid-1980s (equation 4), the response of employment
growth to size was slightly smaller than in the recession periods, suggesting that the tendency for initially larger firms to grow more rapidly than smaller firms was less pronounced. The coefficient on the debt-to-asset ratio bore the opposite in sign from that
in the recession periods; firms with higher initial debt-to-asset ratios grew more rapidly
in the mid-1980s. However, the employment coefficient for the expansion period differed only slightly from that for the recession periods, and the negative debt-to-asset coefficient was statistically indistinguishable from zero.
The model for inventories is similar to the employment model, with the inventoryto-sales ratio as the dependent variable. As Table 9 shows, these patterns have shifted
significantly, but in the opposite direction from that which was hypothesized. Thus, in
the most recent period, we see a greater increase of inventories relative to sales in firms
with high initial debt-to-asset ratios. This result represents a sharp turnaround from previous recession periods, when the relationship between indebtedness and inventory-tosales ratios was negative (and significant in 1972-75). There is no evidence that firms
with larger initial employment better maintained inventory-to-sales ratios in any recession period. The results for the mid-1980s (equation 4) suggest that the switch in sign
on the debt-to-asset ratio occurred about then.
Results for a variable related to investment are shown in Table 10. The dependent
variable is the ratio of net plant to sales. We express the model in terms of the stock of
net plant, rather than the flow of capital spending, because many firms report zero or
negative levels of capital spending. Because the value of net plant can change from depreciation and revaluation and accounting changes, as well as actual spending on new
capital, the connection between these results and capital spending is somewhat loose.
The equations consistently show that firms with high levels of initial employment had
high levels of net plant relative to sales. Thus, these results suggest that smaller firms
either reduced capital spending or shed assets more quickly than did large firms in the
most recent episode.
The most interesting item in Table 10 is the shift in the coefficient on the initial debtto-asset ratio. In 1972-75 and 1979-83 there was a strong positive relationship between
Table 9: Determinants of Inventory Levels in Compustat Sample
388
Initial Employment
Initial Debt-toAsset Ratio
R2
1.1972-75
.005
(.007)
-.338
(.105)
.294
2. 1979-83
-.024
(.006)
-.007
(0.94)
.294
3. 1988-91
-.021
(.008)
..309
(.105)
.355
4. 1984-87
-.046
(.008)
-.028
(.116)
.321
Period
Model:
Log li]t - Log Sijt = a log Eij0 + b(D/A)ij0 + Cj + dt + e.
Ljjt = inventories of firm i of industry j in year t.
Sjjt = sales of firm i of industry j in year t.
Note: Standard errors in parentheses.
initial debt-to-asset ratios and the ratio of net plant to sales; in 1988-91 that relationship
became negative. To the extent that the ratio of net plant to sales is correlated with capital spending, this shift is evidence that high ratios of debt to assets inhibited such spending to an unusual degree in recent years. This effect appears to be a recent phenomenon;
equation 4 shows that the positive association between high debt and high ratios of net
plant-to-sales continued through the mid-1980s.
Finally, we estimate the determinants of research and development spending normalized by sales. The equations of Table 11 show a consistent and growing negative relationship between debt-to-asset ratios and R&D spending. Thus, we have some evidence
that credit constraints as exemplified by initial debt-to-asset ratios were present in the
past few years. Also compared to the larger firms, there is some evidence that smaller
firms responded to weak sales by greater cuts in R&D spending in the 1988-91 recession
Table 10: Determinants of Net Plant in Compustat Sample
Initial Employment
Initial Debt-toAsset Ratio
R2
1. 1972-75
.090
(.008)
.730
(.120)
.276
2. 1979-83
.084
(.007)
.394
(.111)
.259
3. 1988-91
.095
(.011)
-.229
(.139)
.358
4. 1984-87
.032
(.010)
.399
(.151)
.274
Period
Model:
Log NPjjt - Log Sjjt = a log Elj0 + b(D/A)jj0 + Cj + dt + e.
NPjjt = net plant of firm i of industry j in year t.
Note: Standard errors in parentheses.
Table 11: Determinants of Research and Development Expenditures in Compustat Sample
Initial Employment
Initial Debt-toAsset Ratio
R2
1. 1972-75
-.093
(.056)
-1.554
(.793)
.189
2. 1979-83
.125
(.059)
-3.863
(.897)
.206
3. 1988-91
.177
(.079)
-6.343
(1.044)
.305
4. 1984-87
.298
(.075)
-3.688
(1.104)
.261
Period
Model:
Log RDjjt - Log Sijt = a log Ejj0 + b(D/A)ij0 + Cj + dt + e.
RDjjt = research and development expenditures of firm i of industry j in year t.
Note: Standard errors in parentheses.
389
Causes and Consequences
period than in earlier ones. Equation 4 suggests that large firms more aggressively expanded R&D spending during the prosperous mid-1980s than in the recession periods
before and after.
The regression evidence does not suggest that smaller or more indebted firms suffered greater contractions in all dimensions of activity in recent years. In some significant areas, however, such relationships do appear: there is evidence of an unusual
negative connection, starting in the late-1980s, between debt-to-asset ratios and the ratio
of net plant to sales, as well as a continuing positive relationship between firm size and
the net plant-to-sales ratio. A negative relationship between debt-to-asset ratios and the
ratio of R&D spending to sales intensified during the most recent period, while positive
relationships between firm size and employment growth and the ratio of R&D spending
to sales ratio continued. On the whole, then, it does appear that smaller firms experienced some disproportionate losses in activity over the last few years, though not to a
more marked extent than in past recessionary periods. In contrast, highly indebted firms
may well have been more severely affected than in past periods, a finding that is certainly consistent with the hypothesis of significant credit supply effects.29
VI. Summary and Conclusions
390
We have found only tentative evidence of an association between credit market tightness
and weakness in the small firm sector. Credit conditions probably tightened more for
small firms than for large firms over the last few years. We have not, however, been able
to establish that small firm performance has been greatly at variance with historic experience.
The small firm sector did not experience an unusually steep decline in employment
in the 1990-91 recession, but did it recover unusually slowly from the recession. The
divergence from the large firm sector was not large, however. Stratification of manufacturing industries by establishment size uncovers at most very modest evidence that size
was inversely related to job loss, weakness in capital spending, or contractions in inventories in the last few years—whether the weakness in activity is measured absolutely or
scaled, using regression analysis, to cyclical factors or longer term trends. Thus, the relatively greater weakness in bank lending to small manufacturers was not associated with
pronounced relative weakness in activity among these firms. Of course, it is plausible
that different factors were at work in the small and large firm sectors: a bank credit
crunch depressing small firm activity, and the restructuring movement weighing on
large corporations. However, if both of these factors had been significant, the decline in
overall business activity probably would have been larger than observed.
Nevertheless, our results suggest that industry and firm financial debt levels were associated with unusual developments in activity over the last few years. Given demand
conditions, manufacturing industries that ended the 1980s with high ratios of net interest
payments to cash flow tended to contract employment in the 1989-91 period to an unusually large degree, given demand conditions, though they were not necessarily the
29
As indicated in the text, a much smaller sample of firms had meaningful data on capital spending. The set
of firms with usable data on employment, inventories, net plant, R&D spending, and capital spending was
approximately half the size of that using the first four variables as screens. Repeating the regressions
shown in Tables 8-11 with the smaller sample gave generally similar results, except that the coefficient on
debt in the net plant model in the 1988-91 periods was positive (though not significantly different from
zero and smaller in magnitude than in the earlier periods). For capital spending (the dependent variable
was the ratio of capital spending to sales) the coefficient on initial employment was uniformly negative
and that on the debt-to-asset ratio was positive.
leaders in paring jobs in an absolute sense. The Compustat data suggest that R&D
spending and the growth of net plant (but not employment growth) in the last few years
were unusually weak, given sales growth, for firms starting the period with high ratios
of debt to assets.
The evidence can be interpreted to mean that indebted businesses tended to achieve
unusually strong productivity gains (weakness in employment relative to production)
and cut back their acquisition of capital and their R&D budgets. However, these phenomena are not necessarily unusual; earlier periods of economic weakness saw somewhat similar events. Thus, there is little or no clear-cut evidence of an unusually strong
association of debt with retrenchment in activity over the last few years. Our final conclusion, then, is that credit supply problems, to the extent that they manifested themselves in changes in activity at indebted firms, probably played a somewhat more
significant role in the weakness in business activity over the last few years than in past
recessionary periods, but were not the predominant forces at work in the business cycle.
Appendix 1. Impediments to Small Business Lending (by Paul Ludwig)
Among the factors contributing to the recent slowdown in small business lending are
significant changes in bank lending practices. These changes - arising from legislative,
regulatory, and internal industry responses to the weakening of the banking sector in the
late 1980s - have raised the costs of extending and applying for business credit. In general, these costs represent a larger fraction of a small dollar credit; as many businessmen
and bankers have noted,1 small business lending has as a consequence been particularly
affected by the changes. This appendix deals with some of the major structural impediments that have hampered small business lending in recent years.
Increased Use of Real Estate Appraisals
One of the major provisions of the 1989 savings and loan bill, the Financial Institutions
Reform, Recovery and Enforcement Act, requires that a real estate appraisal be conducted for any property involved in a commercial loan transaction exceeding $ 100,000. For
many of these loans, a detailed, fourteen-point evaluation must be completed by an approved, certified appraiser, whether or not the real estate is used as a primary source of
collateral. The net effect of this requirement is to increase the time and expense involved in processing the loan. The requirements are especially burdensome for small
business owners, who typically use real estate to collateralize their loans. With prices
ranging from $2,500 to $10,000, 2 the appraisals represent a significant increase in the
costs of a small business loan.
Increased Loan Documentation Requirements
The amount of documentation required to secure a loan has increased significantly. In
addition to arranging the appraisals mentioned above, borrowers must file several years
of financial statements and tax returns, a business plan, and a set of business projections
within thirty days of applying for the loan. While this requirement is a minor burden for
1
For a collection of such opinions, see "The Credit Cruch For Small- and Medium-Sized Businesses", Field
Hearing before the Committee on Banking, Finance, and Urban Affairs, House of Representatives. March
20, 1993.
2
"Regulatory and Supervisory Impediments to Small Business and Middle Market Lending", Mellon Bank
Corporation, in "The Credit Crunch For Small- and Medium-Sized Businesses", p. 190.
391
Causes and Consequences
larger firms, small businesses often do not have the sophistication to produce such documents in a timely manner. To meet the requirements, borrowers often need to spend a
significant amount of time preparing the paperwork, or they must pay to get outside
help. Although lending officers can extend the credit without complete documentation,
they would then run the risk of having the loan classified by the bank examiners.
Higher Bank Capital Standards
392
To improve the industry's ability to absorb losses, regulators have required banks to increase their capital holdings. These new risk-based capital requirements force banks to
set aside more capital per dollar of commercial loan than is the case for consumer loans
or security holdings. As a result, banks have an incentive to shift their portfolios away
from commercial loans. Moreover, new capital requirements may have especially constrained the growth in small business credit because small firms generally have few alternatives to commercial banks as sources of credit. Larger firms have greater access to
funds raised in securities markets, which are not affected by the capital requirements.
Appendix 2 Table 1: Dependent Variable: Employment Growth
Coefficient
Constant
Production
Growth
R2
Durbin-Watson
Nonelectric
machinery
-0.859
(0.457)
0.296
(0.066)
0.727
1.375
Fabricated metals
-0.503
(0.283)
0.528
(0.049)
0.842
1.755
Electrical machinery
-0.671
(0.303)
0.317
(0.059)
0.746
1.732
Furniture
-0.238
(0.299)
0.388
(0.063)
0.680
1.889
Instruments
-0.275
(0.291)
0.147
(0.067)
0.633
1.431
Lumber
-0.387
(0.388)
0.342
(0.047)
0.660
1.832
Miscellaneous
durables
-0.433
(0.235)
0.232
(0.069)
0.485
1.861
Primary metals
-1.058
(0.493)
0.234
(0.023)
0.786
1.775
Stone, clay, and glass
-0.475
(0.209)
0.282
(0.055)
0.544
1.934
Other transportation
equipment
-0.454
(0.334)
0.314
(0.037)
0.726
1.882
Motor vehicles
-0.585
(0.527)
0.307
(0.030)
0.745
1.818
Apparel
-0.513
(0.212)
0.421
(0.063)
0.680
1.918
Food
-0.139
(0.088)
0.110
(0.075)
0.055
2.048
Chemicals
-0.057
(0.194)
0.047
(0.034)
0.556
1.843
Printing
0.348
(0.286)
0.014
(0.027)
0.764
1.284
Leather
-1.083
(0.353)
0.296
(0.060)
0.604
1.890
Paper
-0.082
(0.163)
0.105
(0.034)
0.525
1.998
Rubber
-0.297
(0.233)
0.454
(0.054)
0.712
1.885
Tobacco
-0.635
(0.287)
0.062
(0.053)
0.055
2.016
Textiles
-0.557
(0.282)
0.231
(0.037)
0.701
1.850
Industry
Petroleum
8
Notes: Standard errors in parentheses. Estimation period: 1980-1 to 1992-11.
a
- Negative coefficient on both current and lagged production variables.
393
Causes and Consequences
Appendix 2 Table 2: Dependent Variable: Log of Real Gross Investment
394
Coefficient
Constant
Log Production
R2
Durbin-Watson
Fabricated metals
8.229
(1.665)
0.801
(0.365)
0.305
0.401
Electrical machinery
12.084
(0.195)
0.681
(0.043)
0.957
1.580
Furniture
-5.190
(2.201)
3.542
(0.489)
0.826
0.819
Instruments
3.699
(0.423)
1.666
(0.093)
0.967
1.366
Lumber
6.351
(1.239)
1.040
(0.277)
0.562
0.638
Miscellaneous durables
-6.894
(5.455)
2.272
(1.186)
0.250
0.287
4.333
(3.530)
1.590
(0.773)
0.278
0.360
9.780
(0.822)
0.558
(0.184)
0.455
0.492
Other transportation
11.220
(0.153)
0.144
(0.153)
0.617
0.566
Apparel
-19.309
(6.707)
5.244
(1.472)
0.536
0.668
Petroleum
-1.277
(5.301)
2.421
(1.154)
0.286
0.188
Chemicals
-3.343
(2.233)
3.201
(0.511)
0.781
0.717
Printing
8.290
(0.480)
0.795
(0.107)
0.834
0.639
Paper
-2.027
(1.424)
3.135
(0.314)
0.901
1.226
Rubber
7.200
(0.359)
0.555
(0.081)
0.811
0.788
Industry
Nonelectric machinery8
Primary metalsa
Stone, clay, and glass
Motor vehicles
Food
3
Leather
8
Tobaccoa
Textiles8
Notes: Standard errors in parentheses. Estimation period: 1980-91.
a
Negative coefficient on both current and lagged production variables.
Appendix 2 Table 3: Dependent Variable: Real Inventory/Sales Ratio
Coefficient
Constant
Time
R2
Durbin-Watson
Primary metals
11.673
(1.724)
-0.017
(0.003)
0.378
0.198
Fabricated metals
7.621
(0.573)
-0.01
(0.001)
0.658
0.524
Nonelectrical
machinery
24.418
(1.339)
•0.041
(0.002)
0.85
0.17
Electrical machinery
8.309
(0.613)
-0.011
(0.001)
0.666
0.249
Motor vehicles
6.169
(0.649)
-0.01
(0.001)
0.584
0.617
Other transportation
equipment
-3.611
(1.836)
0.014
(0.003)
0.247
0.211
Other durable goods
6.746
(0.476)
-0.009
(0.001)
0.682
0.38
Food
4.109
(0.088)
-0.006
(0.0002)
0.964
0.6
Paper
2.277
(0.387)
-0.002
(0.001)
0.114
0.246
Chemicals
3.814
(0.387)
-0.004
(0.001)
0.403
0.478
Petroleum
2.233
(0.454)
-0.003
(0.001)
0.164
0.23
Rubber
8.312
(0.618)
-0.013
(0.001)
0.721
0.304
Industry
Notes: Standard errors in parentheses. Estimation period: 1980-1 to 1992-11
395
Causes and Consequences
References
396
Board of Governors of the Federal Reserve System. "Credit Availability for Small Businesses and Small Farms." Mimeo, December 31, 1992.
Boldin, Michael. "The Macroeconomic Implications of Increasing Corporate Debt."
Federal Reserve Bank of New York, February 1992.
Brauer, David. "A Historical Perspective on the 1989-92 Slow Growth Period." Federal Reserve Bank of New York Quarterly Review, vol. 18, no. 1 (Summer 1993),
pp. 1-14.
Cantor, Richard. "A Panel Study of the Effects of Leverage on Investment and Employment." In Studies on Financial Changes and the Transmission of Monetary Policy,
Federal Reserve Bank of New York (1990), pp. 119-134.
Dunkelberg, William C , and William J. Dennis, Jr. "The Small Business 'Credit
Crunch.1" NFIB Foundation, December 1992.
Fazzari, Steven M., R. Glenn Hubbard, and Bruce C. Petersen. "Financing Constraints
and Corporate Investment." Brookings Papers on Economic Activity, 1988:1,
pp. 141-195.
Gertler, Mark, and Simon Gilchrist. "Monetary Policy, Business Cycles and the Behavior of Small Manufacturing Firms." Federal Reserve Board, Finance and Economics
Discussion Series no. 93-4, February 1993.
Radecki, Lawrence J. "A Review of Credit Measures as a Policy Variable." In Intermediate Targets and Indicators for Monetary Policy: A Critical Survey, Federal
Reserve Bank of New York (1990), pp. 183-231.
The Credit Slowdown and the Monetary
Aggregates
by R.S. Hilton and C.S. Lown]
Over the past three years, a sharp decline in credit growth has been accompanied by extreme weakness in the broad monetary aggregates. The behavior of these measures is
not unusual insofar as money and credit—especially credit at depository institutions—
have generally moved together during the past thirty years (Charts 1 and 2). However,
the very sharp deceleration in money and credit growth over the last few years raises the
question of how these aggregates have been affected by special factors during this most
recent business cycle.
In this study, we assess the recent slowdown in the broad measures of money by analyzing several factors that are believed to have affected the supply of and demand for
both money and credit, including weakened bank balance sheets, a reduced willingness
to lend, shifts in consumers' preferences for money, and shifts in borrowers' demand for
loans. In the first section of the paper, we lay out a framework for discussing the various
channels connecting credit and money and describe how the principal forces behind the
slowdown in credit formation may have affected the monetary aggregates. In the second section, we estimate the impact of the credit slowdown during the past few years on
M2, controlling for business cycle influences. An econometric analysis of the links between the credit and money slowdown is undertaken in the final section. In that section,
the predictive performance of M2 demand equations over the past few years is evaluated, the causal relation between money and credit is examined, and the ability of various
bank balance sheet measures to explain deposit formation is studied in a cross-sectional
framework.
The evidence reviewed in this study suggests that recent growth in the monetary aggregates has been weakened by several factors that have slowed credit formation—factors operating in addition to the direct effects associated with the 1990-91 recession and
subsequent recovery period. These factors include a supply-side credit crunch in the depository sector and a retrenchment in credit demand by heavily indebted borrowers. A
shift in consumer preferences towards holding more nonmonetary assets also has led di1
We thank Joe Abate and Joe LaVorgna for their research assistance and Akbar Akhtar for helpful comments.
397
Causes and Consequences
Chart 1: Total Nonf inancial Private Credit, M2, and M3
Four Quarter Growth Rates
Percent
20
1960
64
68
72
76
84
88
92
76
84
88
92
Note: Shaded areas represent recessions.
Chart 2: Depository Credit, M2, and M3
Four Quarter Growth Rates
398
Percent
20
1960
64
68
72
Note: Shaded areas represent recessions.
rectly to some weakness in the broader monetary aggregates and indirectly may have
contributed to slower bank credit formation. The behavior of interest rates, with bank
loan rates rising relative to market yields and deposit rates falling relative to alternative
yields, suggests that a contraction in the supply of bank loans has been an important
force weakening both credit formation and the broader monetary aggregates. Simple
calculations suggest that, combined, all the factors cited above could have depressed the
level of M2 by roughly $300 billion, or nearly 10 percent, by the third quarter of 1992,
two years after the onset of the most recent economic downturn.
An evaluation of the forecast performance of a standard M2 demand equation yields
a similar-sized value for the cumulative prediction error for this aggregate in the third
quarter of 1992. Including various measures of credit creation and alternative measures
of the opportunity cost of holding M2 deposits significantly improves the statistical fit
of the estimated M2 equation for recent years. Results from Granger-causality tests provide some weak support for the view that the causal link running from credit to M2 money creation has strengthened in recent years, a finding consistent with the widespread
view that depositories have increased their use of smaller deposits (a component of M2)
as a managed liability to fund loans. Meanwhile, little evidence is found of a statistically significant causal link running from money to credit for the recent period. Finally, a
number of cross-section results performed using state-level and bank-level data show
that measures of bank balance sheet distress account for at least a modest portion of the
weakness in deposit formation in recent years.
Overall, the empirical work undertaken in this study shows that factors that have
shifted credit demand and supply, and money demand—apart from direct business cycle
influences associated with the 1990-91 economic downturn—account for a considerable
portion of the weakness in M2 seen in recent years. We are not able to isolate fully the
effect of credit supply problems on the monetary aggregates in our empirical analysis.
However, much of the evidence found—from examining the behavior of key interest
rate spreads, by including credit supply measures in the estimation of M2, and the crosssection estimation results—supports the view that credit supply constraints played a significant role in weakening the broader monetary aggregates.
Credit Formation and the Monetary Aggregates
In this section, we begin by outlining the linkages between credit formation and the
monetary aggregates, and then turn to the specific reasons commonly cited for the slowdown in these aggregates.
Links Between Credit Formation and the Monetary Aggregates
The channels between credit formation and the monetary aggregates are multiple and
complex. Nonetheless, the key linkages between credit and money may be summarized
in the framework of supply and demand schedules for depository lending and deposit
creation. Depository institutions, including banks, act as a principal source of credit for
many businesses and most households. Depositories also supply a large fraction of the
liabilities that are included in the M2 and M3 monetary aggregates.
Stylized representations of the supply and demand for bank lending and money deposits are presented in Chart 3. In the top panel, bank loan supply and demand are
drawn as a function of the spread between the bank loan rate (rp) and the market yield
associated with investment or borrowing alternatives (r), represented by a single rate
that is held exogenous in this simple framework. Similarly, the supply and demand for
399
Causes and Consequences
money deposits at banking institutions are determined in part by the opportunity cost of
holding or creating these deposits, represented in the bottom panel by the spread between the deposit rate (rd) and the alternative market yield.2
Factors other than lending and deposit rates can also affect the supply and demand
for either bank lending or monetary liabilities. For example, lower aggregate income
would decrease the demand for both bank credit and deposits, moving their respective
demand schedules to the left. Furthermore, some developments that bear directly on aggregate measures included in the consolidated balance sheet of depository institutions
may spark a response by banks designed to keep their assets and liabilities in overall balance and at desired levels. Such responses may appear as shifts in the bank lending or
deposit supply schedules. For instance, developments that reduce the willingness of depositories to extend credit, moving the supply schedule of loans to the left, would also
reduce banks1 funding needs, shifting the money supply schedule to the left as well.
The degree to which a change in the supply or demand for bank lending might translate into a shift in the money supply schedule depends importantly on several factors,
including (1) how extensively the managed liabilities in the monetary aggregate under
consideration are used to finance lending, and (2) how the investment portfolios of
banks are managed. These two factors are discussed in more detail below. Also discussed below is how total credit responds to factors that affect bank lending.
2
The alternative market rates used to measure the opportunity costs of borrowing and lending and of holding or creating money deposits are exogenous in this framework, so a single rate can be used in the graphical treatment to represent all market rates with no loss of generality.
Chart 3: Bank Lending and Money
400
Deposits
Loans
r-rd
Ls
= supply schedule of depository loans.
LD
= demand schedule for depository loans.
Ms
MD
(rp-r)
(r-rd)
L
M
=
=
=
=
=
=
-
supply schedule of monetary deposits.
demand schedule for monetary deposits.
spread between the bank lending rate (rp) and the market rate (r).
spread between the market rate (r) and the bank offering rate (rd).
level of depository loans.
level of deposit balances.
Managed liabilities in the monetary aggregates
One factor that influences the degree to which a change in bank lending affects money
is the extent to which the monetary aggregate under consideration is used to fund lending activities. For many years, large certificates of deposits—those worth $100,000 or
more and included in the current definition of M3—had been used as the primary tool
by banks to finance their lending on the margin. Banks would adjust yields offered on
these deposits (in effect, shifting the supply schedule for M3) to attract the funding
needed to support their desired lending. As a result of financial deregulation that has
removed most deposit-rate ceilings, during the past several years depositories have increasingly relied on small time deposits—those under $100,000 in value and included
in the current definition of M2—to help meet their funding needs. This shift in bank
practices has broadened the link between bank lending and M2; thus, shifts in bank
credit supply or demand are now more likely to lead to significant changes in the M2
deposit supply function than was previously the case.3
Holdings of government securities
The degree to which depositories offset changes in lending supply or demand by shifting their other asset holdings will also partly determine the response of the money supply schedule to lending changes.4 The top panel of Chart 4 indicates that a significant
portion of the slowdown in depository lending in the past few years has been partly offset by more rapid accumulation of investment securities.5 Nonetheless, overall depository credit, which includes these securities, has fallen considerably in recent years
(bottom panel). Thus, the overall balance sheet constraint suggests that a substantial
part of the reduction in depository lending has been met with a decline in bank deposits.
Reliance on banks as a source of credit
This study focuses primarily on the association between depository credit formation and
the monetary aggregates, which are used to finance a major portion of bank lending.
Some general observations about the relation between changes in bank credit formation
and the behavior of total credit should be made. A decline in depository credit that is
accompanied by a reduction in money may not be reflected in a similar change in total
credit formation because businesses and households have some ability to raise credit
outside of banking channels. Many of the financial innovations of the past decade have
made it easier for larger corporations to borrow directly in credit markets or through
nondepository institutions. Consequently, a shift in the supply of depository credit will
not necessarily be fully reflected in changes in total credit.6 Nonetheless, many small
and medium-sized firms remain heavily reliant on depository institutions to meet their
3
Wenninger and Partlan (1992) found that the contemporaneous correlation between interest rates on small
and large time deposits, and between the growth of these two types of deposits, was significantly higher
over the sample period running from the late-1970s to the present than it was over the preceding twenty
years. The increased correlation in the behavior of these deposit categories suggests that banks now use
some M2 deposits as they traditionally have used M3 deposits to fund credit expansion.
4
The rate of return on these alternative assets is captured in the market interest rate presented in the graphs.
Thus, the sensitivity of lending or deposit creation to changes in security holdings is reflected in the supply
elasticities.
5
In fact, some anecdotal evidence suggests that stricter capital requirements have encouraged banks to substitute government securities in place of loans.
6
The effects of a slowing in depository lending on economic activity, which are not treated explicitly in this
study, may depend in part on how readily nonbank credit can be substituted for bank lending.
401
Causes and Consequences
Chart 4A: Depository Lending and Investments
Four Quarter Growth Rates
Percent
40
1960
64
68
72
76
80
84
Note: Shaded areas represent recessions.
Chart 4B: Total Credit of Depository Institutions
Four Quarter Growth Rates
402
1960
64
68
72
Note: Shaded areas represent recessions.
76
84
88
92
credit demands, so a fall in depository credit supply, and the associated drop in money
supply, will at least be partly reflected in a decline in overall credit formation.
Factors Behind the Credit Slowdown and Implications for Money Growth
The simple framework presented above can be used to illustrate the consequences of a
credit slowdown for the monetary aggregates. A decreased willingness on the part of
depository institutions to lend is widely regarded as a principal cause of the slowdown
in depository credit formation.7 A "supply-side" credit crunch has been tied to several
related developments. Reduced depository credit supply has been attributed to a shrinking capital base that directly impedes the ability of depositories to lend, and to loan losses accumulated over the past several years.8 Stricter capital requirements and more
stringent lending guidelines imposed by banking regulators in response to these same
factors have reinforced banks' reluctance to lend. The widespread distress experienced
in the savings and loan industry and the associated collapse in the capital base of thrifts
is one part—albeit a substantial one—of this credit crunch phenomenon.
All these developments would shift the supply schedule for bank loans, as portrayed
in Chart 5A, placing upward pressure on bank lending rates measured relative to market
interest rates. At the same time, after allowing for any substitution of investment securities in place of direct lending, the decreased loan supply would contribute to a leftward
shift in the money supply schedule as the funding needs of depositories fell, also shown
in Chart 5A. As a result, the money supply would be reduced, and bank offering rates
would fall relative to competing market yields. In this scenario, bank lending rates rise
relative to market rates while deposit rates fall, and the amount of outstanding bank
loans and deposits falls.
Considerable anecdotal evidence suggests that banks have relied in part on nonprice
mechanisms to allocate credit supplies. To ration credit, depositories can adjust a variety of qualifying standards. Certain forms of collateral may be revalued to reflect revised perceptions about underlying risk, or credit may be restricted to customers with
established, long-term relationships. In recent years, many borrowers in the real estate
sector in particular have reported being effectively shut out of the bank credit market.
In terms of the graphical framework presented in Chart 5 A, under credit rationing, lending rates are held below market clearing levels, where loan demand outstrips supply, and
banks cut back on desired lending through nonprice means. While potentially significant, large reductions in desired lending are difficult to achieve entirely through nonprice means, and some combination of higher lending rates and more stringent lending
criteria are likely to have been used.
The credit slowdown has also arisen partly from "demand-side" factors not directly
linked to the aggregate level of economic activity as businesses and households have
voluntarily scaled back their borrowing. In many cases, the high levels of debt incurred
in the 1980s and depressed values of asset holdings, particularly real estate, have left
borrowers reluctant or financially unable to take on additional debt. A falloff in credit
demand would tend to reduce bank lending rates, as shown in Chart 5B. This decline
in demand would also contribute to a leftward shift in the money supply, also portrayed
in Chart 5B, because the funding needs of depositories would be reduced. In this case,
7
Explanations for the weakness in lending by depository institutions arc discussed in greater detail by Lown
and Wenninger (1993).
8
Peck and Roscngren (1992) have argued that a capital shrinkage has constrained bank lending in the New
England region. Johnson (1992) makes a similar case at the national level.
403
Causes and Consequences
Chart 5: Shifts in Bank Lending and Money*
404
A: Decline in Bank Lending Supply
Loans
rp-r
B: Decline in Credit Demand
Loans
Deposits
M:
C: Decline in Money Demand
Loans
Deposits
M:
* See notes after Chart 3.
both bank lending and money decline, and both bank lending rates and offering rates on
deposits fall relative to competing market yields.
Factors other than a slowing in bank credit demand and supply have been linked to
recent weakness in money growth. In particular, a shift in consumer preferences for
holding money balances has been tied to the introduction of alternative investment vehicles offered by nondepository institutions and may have been a separate source of
weakness in money demand. Bond and stock mutual funds have found increasing acceptance among smaller investors in recent years. These accounts often carry marketbased returns and offer a degree of liquidity similar to competing deposits included in
the broader monetary aggregates. Their introduction would contribute to a leftward
shift in the money demand schedule as consumer preferences change, as represented in
Chart 5C, putting upward pressure on deposit rates.9 The massive closing of thrifts by
official regulators also reportedly has encouraged a reallocation of funds into accounts
not included in the monetary aggregates that might not otherwise have taken place,
prompting a shift in money demand. A fall in money demand impairs the ability of depositories to attract funds without raising offering rates, causing credit supply to be cut
back, also shown in Chart 5C. I() In this case, money and credit decline, while both bank
lending and offering rates rise relative to competing market rates.
Finally, the broad decline in aggregate economic activity associated with the 199091 recession, as well as the sluggish pace of growth that marked the early stages of the
subsequent recovery, have undoubtedly exerted a major influence on the behavior of
credit and money.11 A decline in economic growth typically weakens credit demand,
including demand for depository lending, and at the same time directly reduces money
demand. Meanwhile, the capital positions of depositories and the value of their collateral holdings normally deteriorate in the face of a slowing economy, reducing the willingness of banks to lend and producing leftward shifts in the supply of loans and
money.12 The net result is reductions in the supply and demand for money and in bank
lending, with an uncertain impact on interest rate spreads.
The impact of the factors cited in this section on interest rate spreads and the level
of bank lending and the money supply are summarized in Table 1. It may be difficult to
establish with any certainty which of these developments have been most responsible
for the sustained weakness in depository credit formation and the monetary aggregates
over the past few years because all of these factors would tend to work in the direction
9
10
1
A heightened consumer awareness of the availability of alternative investment opportunities would also
increase the elasticity of demand for money to movements in competing market interest rates.
Viewed from a longer term perspective, the growth in money market mutual funds and alternative investment accounts may be seen as part of the ongoing trend toward greater disintcrmediation in credit markets,
which has included an expanded role for nondepository institutions. This process would be associated
with a decline in the demand for bank lending as well as a shift in money demand, exerting pressure on
bank profit margins measured by the spread between bank lending and offering rates. Over time, this pressure on bank earning spreads would force some depositories to fail, shrinking the entire sector and causing
the supply schedules for bank credit and deposits to shift leftward (thereby restoring bank earning spreads
to original levels).
' To an uncertain degree, recent economic weakness may have been brought on, or at least been reinforced
by, some of the interruptions to credit supply and demand described above; however, the weakness in economic activity over this period is treated as having had an independent effect on money and credit in this
analysis.
12
In terms of the graphical framework of Chart 3, ultimately, the supply and demand schedules in both the
bank lending and deposit markets would shift to the left, reducing the quantities of lending and money, but
with uncertain implications for interest rate spreads.
405
Causes and Consequences
of weakening credit formation and the monetary aggregates. However, these factors
hold different implications for the behavior of interest rates spreads, so an examination
of these yields might provide some idea of which forces have been dominant.
Bank lending rates (measured by the prime rate) have not declined as much as other
borrowing rates (represented by the commercial paper rate) since the onset of the last
recession (Chart 6). This behavior suggests that constraints on bank credit supply may
have been a more important factor in the slowing of depository credit formation—and
therefore of the monetary aggregates—than a slowing in business and household credit
demand tied to balance sheet restructuring efforts. To be sure, a widening in this spread
could also be associated with a leftward shift in money demand, as portrayed in Chart
5C. However, the simultaneous widening in the spread between six-month bank CD
rates and Treasury yields of comparable maturity, shown in Chart 7, suggests that a drop
in credit supply rather than a decline in money demand better explains the recent behavior of these aggregates and related interest rate spreads.
These general inferences must be treated cautiously on the basis of the circumstantial
evidence presented. All of the economic developments described above, as well as others, have undoubtedly affected the supply and demand schedules for credit and money
in recent years. In the face of so many simultaneous influences, simple observations of
interest rate spreads could give an incomplete picture of the primary forces affecting the
behavior of the monetary aggregates and credit.13
In the remainder of this paper, we attempt to quantify how much of the recent weakness in the broader aggregates can be directly linked to the slowing in credit formation.
Rough measures of the effects of all factors other than direct business cycle influences
on M2 are made in the following section. No attempt is made to sort out the relative
contributions that supply versus demand influences on credit formation have had on
money. Further work in the subsequent section, however, effectively controls for some
of the demand-side factors that may have weakened credit formation, providing a better
sense of the importance of supply-side lending constraints for the monetary aggregates.
13
Feinman and Porter (1992) suggest that the standard measures of bank offering rates may be flawed. Their
work indicates that better measures of yields on nontransactions deposits in M2 show an even more
marked decline relative to other market interest rates than is shown in Chart 7.
Table 1: Effects of Exogenous Shocks on Interest Rate Spreads,
Money and Depository Lending
406
Spread Between:
Bank Lending Rate
and Market Rate
rp-r
Market Rate and
Deposit Rate
r-rd
Money
M
Depository
Loans
L
Decreased loan supply
+
+
-
-
Decreased loan demand
-
+
-
-
Decreased money
demand
+
-
-
-
Exogenous Shocks
Chart 6: Prime Rate less Six-Month Commercial Paper Rate
Percent
1959
62
65
68
71
74
77
80
83
86
89
92
Note: Shaded areas represent recessions.
Chart 7: Spread between the Six-Month Treasury Bill Rate and the Six-Month CD Rate
1959
62
65
68
71
74
77
83
86
89
92
Note: Shaded areas represent recessions.
407
Causes and Consequences
II. Measuring the Effects of the Recent Credit Slowdown on M2
408
In the previous section, we considered how changes in loan supply and demand, money
demand shifts, and cyclical factors may have affected bank credit and deposits. Over
the last few years, all of these factors have reinforced one another and slowed the growth
of bank credit and the broader monetary aggregates. In this section, two independently
derived measures are taken of the net impact that factors apart from the business cycle
have had on M2 since the onset of the last recession. An alternative calculation attempts
to control for the effect that the secular contraction of the depository sector has had on
money.
By comparing the recent behavior of credit and the monetary aggregates to their historical patterns around previous business cycles, some measure can be made of the effects that all noncyclical forces have had on M2 in the past few years, including
reductions in the supply and demand for bank credit.14 Of course, this approach is imprecise, and does not allow for the indirect effects some of the sources of weakness in
credit could have had on the money and credit aggregates by depressing economic activity. This method also may not fully control for how the behavior of credit and money
responds to economic downturns of different intensity.
The behavior of credit and money relative to the level of nominal income since the
onset of the last recession (1990-IH) demonstrates that extraordinary economic forces
have operated to weaken these aggregates, even allowing for the direct effects of the
economic downturn (Chart 8). 15 Depository credit measured relative to GDP began
edging down in the quarters leading up to the onset of the most recent recession, but
compared to past business cycles it weakened markedly only after the recession was underway. At the same time, the broader monetary aggregates measured relative to nominal income also began to weaken measurably using past experience as a benchmark.
Many of the factors cited in the previous section—a depository credit crunch, a retrenchment in credit demand not directly related to the recession, and a shift in preferences for holding nonmonetary liabilities—were believed to have been operating at this
time to weaken money and credit.
Recent values of key measures of credit and money are presented in Table 2. In the
two years since the start of the last economic downturn—from 1990-III through 1992III—the value of M2 as a percentage of nominal GDP, on balance, fell by nearly 3 percentage points. This behavior stands in contrast to past experiences around business cycles. As shown in Chart 8C, M2 relative to nominal GDP rose, on average, about
4 1/2 percentage points in the two years following the onset of previous recessions. Had
M2 conformed to this pattern during the latest business cycle, its value would have been
about $270 billion higher in 1992-III than it was. 16
An alternative estimate is derived by calculating the dollar value of depository credit,
14
The behavior of bank credit rather than bank lending is examined in this section because some of the weakness in lending has been met with an increase in holding securities rather than a decline in deposit creation.
15
Wenninger and Partlan show that the weakness in M2 and M3 over this time was concentrated in the small
and large time deposit components. A weakening in the monetary aggregates associated with a broad
decline in demand stemming from an economic slowdown might be expected to have been more evenly
distributed across the components of the monetary aggregates.
16
This calculation simply raises the level of M2 in the eighth quarter following the beginning of the last
recession--1992-111-by enough to bring the dashed line in the bottom right panel of Chart 10 to the level
of the solid line.
Chart 8A: Ratio of Total Private Credit to GDP
Indexed to Recession Expansion
-4
-3
-2
-1
P
1
2
3
4
5
6
7
8
Note: Average is the average of the 1960, 1969, 1973, 1nd 1981 recessions.
Chart 8B: Ratio of Depository Credit to GDP
Indexed to Recession Expansion
Index
1 IU
105 - ^ ^ ^ ^
Average
^nuinuui
100
95
^ ^ ^
'
i
^^^8I»
-
^"^*V^^
90
" ^ mmnmr"
"
1
miniiiiiiiiiiiiiW' '
89Q4-92Q4
-
-4
-3
-2
-1
P
i
i
i
i
i
i
i
i
i
1
2
3
4
5
6
7
8
9
Note: Average is the average of the 1960, 1969, 1973, 1nd 1981 recessions.
409
Causes and Consequences
Chart 8C: Ratio of M2 to GDP
Indexed to Recession Expansion
Index
y
104
-
—
• •—l
/
102
•—
Average"
mm^^
w
100
>.
98
-
96
-
•
i
i
i
-4
-3
-2
-1
QA
P
89Q4 - 92Q4
1
1
1
1
1
1
1
1
1
2
3
4
5
6
7
8
Note: Average is the average of the 1960,1969, 1973, and 1981 recessions.
Chart 8D: Ratio of M3 to GDP
Indexed to Recession Expansion
410
Index
108
-4
-3
-2
-1
P
1
2
3
4
5
6
7
Note: Average is the average of the 1960,1969, 1973, and 1981 recessions.
8
1
and associated amount of funding with M2 deposits, that would have been necessary to
keep the ratio of depository credit to nominal GDP in this recession at its average level
during past recessions (taking the level of nominal income as given). In the first eight
quarters after the start of the last economic downturn, depository credit relative to income fell roughly 10 percentage points, whereas in past episodes this ratio rose by about
3 1/2 percentage points in the same length of time, as shown in Chart 8B. If a higher,
counterfactual ratio of depository credit to income in line with past experience had been
financed with a "typical" mix of monetary and nonmonetary depository liabilities, then
the nontransactions components of M2 at depository institutions would have been about
$320 billion (i.e., about 15 percent) higher as of 1992-111.
Each of the above calculations attempts to measure the effect that all factors not directly related to the economic recession have had on M2, but by different means. The
estimates obtained, although diverse, are of a similar order of magnitude and suggest
that extraordinary factors—including shifts in credit supply and demand and in money
demand—had depressed the level of M2 between 7 1/2 to 9 1/4 percent as of 1992-HI.17
An alternative approach attempts to control for the secular downsizing of the depository sector as well as for the economic recession in measuring the consequences of the
credit slowdown for M2. Depositories' share of total private credit creation has been on
17
The values obtained here arc similar in magnitude to the total prediction error for M2 found using a standard econometric model for that aggregate reported in section III.
Table 2: Values of Credit and Money
In Billions of Dollars
1989-1V
1990-111
1992-111
783
812
969
Currency
222
240
283
Demand deposits
278
280
323
Other checkable deposits
283
292
363
Savings deposits and MMDAs
885
920
1146
Small time deposits
1146
1155
927
Nontransaction M2 at depositories
2031
2075
2073
Actual M2
3207
3322
3476
M1
Estimated impact on M2 of:
Restoring the ratio of M2 to GDP to
"normal" levels
+268
Restoring the ratio of depository
credit to GDP to "normal" levels
+317
Restoring the ratio of private credit to
GDP to "normal" levels
Nominal GDP
+93
5340
5560
5979
411
Causes and Consequences
a gradual, but steady, downtrend for almost two decades. The recent expansion of nonmonetary consumer investment funds described in the previous section is partly the
byproduct of the growth of nondepository credit intermediaries, which has contributed
to the shrinkage of the depository sector as a source of credit formation. To control for
this additional factor, movements in the ratio of total private credit to nominal GDP
around past business cycles are used as the benchmark for calculating alternative levels
of depository credit formation and M2 deposit creation.
As shown in Figure 8A, in the first two years since the beginning of the last recession, total credit relative to income fell roughly 2 percentage points, underperforming
by about 4 1/2 percentage points its usual behavior in the same time frame in past business cycles. Under the assumption that this discrepancy measures the weakness in depository credit that can be explained either by the recession directly or by the secular
contraction of the depository sector, we raise the nontransactions components of M2 by
this same percentage amount. This adjustment adds nearly $100 billion to the value of
M2in 1992-IH.
This last calculation must be viewed very tentatively, as it could misstate the effects
on M2 growth of factors not related to the recession and the shrinkage of the depository
sector for several reasons. On the one hand, the effects of a supply-side slowing in depository credit formation on M2 would not be captured to the extent that borrowers were
able to raise credit outside of the depository sector.18 On the other hand, the effects of
a decline in consumer demand for M2 balances, which might be viewed as a part of the
ongoing downsizing of the depository sector, would be captured in the calculation to the
extent that borrowers could not secure alternative sources of funding outside the depository sector.19
The calculations presented in this section are only intended to provide crude approximations of the impact that the credit slowdown has had on the broader monetary aggregates. The methods used were simple, and alternative calculations obtained using
different techniques or assumptions could well yield different results. To examine in
greater detail the relations between depository credit and the broader monetary aggregates, we employ several econometric methods, described in the following section.
III. Empirical Analysis of the Link Between the Credit and Money Slowdown
412
In this section, we use econometric techniques to study several aspects of the relation
between depository credit and the broad monetary aggregates. We first consider the extent to which the recent performance of the M2 aggregate has been unusual by examining its behavior using the money demand model developed at the Board of Governors.
Next, using Granger "causality" tests, we explore whether the slowdown in credit
growth occurred prior to that of money growth. (Finding such a result does not imply
that the lending slowdown caused the money growth slowdown, but it does imply that
the data is at least consistent with this view.) The causality tests will provide some insight into whether the relation between credit and the broader monetary aggregates has
undergone any shift over the past decade as a result of innovations in the financial sector.
Therefore, these tests will suggest whether shifts in credit supply or demand now have
a larger effect on the monetary aggregates in recent years than previously. Finally, using
18
In this case, measured total credit would not have fallen, even though depository credit, and presumably
the level of M2 funding, would have declined.
19
In this situation, total credit would fall along with depository credit and M2.
state- and bank-level data, we conduct a cross-sectional regression analysis to determine
whether banking sector difficulties, proxied by bank balance sheet measures, can explain the money growth slowdown.
M2 Demand Equation
In order to determine how much of the weakness in M 2 has been "unusual," we estimate
the Board of Governors' M 2 demand model and examine the residuals from the estimated equation. Large residuals over the past few years would indicate that the behavior of
M 2 has been unusual relative to its past behavior. These errors capture deviations in
money demand, but because supply factors are not explicitly controlled for, the errors
may also reflect money supply disturbances linked to shifts in credit demand and supply.
The Board's M 2 equation is specified in an error-correction framework, which takes
into account the long-run relationship between M 2 , its opportunity cost, and economic
activity. The dependent variable in the equation is the growth rate of nominal M 2 . In
addition to the variables included in the error-correction specification, the independent
variables include a lagged dependent variable, a time trend, and four dummy variables
to take into account the introduction of nationwide super N O W and M M D A accounts.
We estimate the equation using quarterly data over the period 1964-1V to 1992-III. The
results appear in Table 3.
The lagged levels of the (logarithm) opportunity cost and of velocity, along with the
time trend, represent the difference between the actual and long-run demand for M 2 . As
the table shows, these variables all are significant in explaining the growth in M 2 . The
change in consumption, the opportunity cost measure and the lagged dependent variable
capture the short-run dynamics and are all significant in explaining M 2 growth.
To determine whether this equation accurately represents the behavior of money demand over the entire time period, we examined the estimated equation for structural
breaks. We considered whether the equation estimated prior to the third quarter of 1976
was significantly different from the equation estimated since then and repeated this
same exercise breaking at 1979-III and at 1982-1V. We found no significant break in the
equation at any of these points. Finally, we estimated the equation through 1989-1V and
tested whether the M 2 prediction errors from 1990-1 through 1992-III were large, which
would suggest that a break occurred more recently. Such a break did occur in 1990,
which coincided with the sharp slowdown in bank credit.
A rough sense of the size of this recent change in the behavior of M 2 can be seen in
Chart 9. The figure plots the actual growth rate of M 2 over the 1964-92 time period,
along with the out-of-sample dynamic forecast for the 1990-1 to 1992-111 period based
on the equation estimated through 1989. The predicted values over the forecast period
are much larger than the actual data, indicating that the previous relationship among
M2, economic activity, and the opportunity cost of holding M 2 broke down over the last
few years.
A complementary way to explore the recent behavior of M 2 is to examine the insample residuals from the M 2 equation estimated over the entire time period. As Chart
10 shows, beginning in 1990, the actual minus predicted values of M 2 growth became
negative and relatively large. There were other periods in earlier years when M 2 growth
was weaker than the equation predicted, but to a lesser degree than over the last few
years. Moreover, the recent weakness occurred at a time when monetary policy was
considered expansionary. During or following each of the earlier recessions, when the
Federal Reserve eased policy to boost economic growth, growth in M 2 was generally
stronger than estimated by the equation.
413
Causes and Consequences
To determine whether the unexplained weakness in M2 growth could be accounted
for by the credit growth slowdown, we extended the independent variables in the money
demand equation to include variables which reflect credit market conditions. The variables we chose are of two types: bank balance sheet variables, such as credit aggregates
and the capital stock, and interest rates on competing financial instruments. The inclusion of credit variables in the money demand specification, because of the obvious endogeneity issue, should not be considered an attempt to improve upon the equation's
specification. Nevertheless, such an extension allows us to determine in a rough way
whether the credit growth slowdown directly contributed to the unexplained weakness
in M2 growth. The inclusion of alternative interest rates is not subject to the same criticism. However, the importance of these rates could be explained by developments other than the credit slowdown.
The bank balance sheet variables we consider are depository credit and loans, and
commercial bank credit and loans. These variables allow us to consider whether the explanatory power of commercial banks differs from that of thrifts, and whether the role
Table 3: M2 Demand Equation
Aln(M2) =
,
p2*ln(OpportunilyCost(-l))
0
+ p3*ln(Velocity(-l)) + p4*AIn(Consumption) + P5*Aln(Consumption(-l))
+ P6*Aln(Consumption(-2)) + P7*Aln(Opportunity Cost) + P8*DCON
+ P9*DUM1 + P|0*DUM2 + p n *Aln(M2(-l)
Estimation Period: 1964-IV to 1992-111
414
Coefficient
Estimate
S.E.
t-stat.
Constant
-.050
0.12
4.27
Po
-.000
.000
Pi
.002
P2
Estimate
S.E.
t-stat.
Pe
.063
.064
.99
2.57
P7
-.007
.002
4.10
.002
.90
Pa
-.006
.006
1.11
-.006
.001
4.25
P9
.029
.005
5.61
P3
-.117
.026
4.43
P10
-.010
.006
1.65
P4
.262
.070
3.73
P11
.615
.077
7.96
P5
.060
.071
.84
Adj. R2
Durbin-h
St. Er.
Coefficient
.68
.86
.005
Notes: M2 = nominal M2 money aggregate. DMMDA = dummy variable for the introduction of
MMDAs, 0 through 1982-1V and 1 thereafter. Opp. Cost = the three-month Treasury bill rate
minus a weighted average of the interest rates on the M2 deposit categories. Velocity = GDP/
M2, Consumption = personal consumption expenditures. DCON = dummy variable for credit
controls, 1 in 1980-11 and 0 otherwise. DUM1, DUM2 = short-run dummies for the introduction
of MMDAs, 1 in 1983-1, 0 otherwise, and 1 in 1983-11 and 0 otherwise, respectively.
Chart 9: Actual and Predicted Growth Rates of M2
1965
68
86
92
89
Source: Board of Governor's Model.
Note: Shaded areas represent recessions.
Chart 10: M2 Demand Residuals
Percent
1
1965
89
92
Source: Board of Governor's Model.
Note: Shaded areas represent recessions.
415
Causes and Consequences
of loans is different from that of security holdings. We also include the capital-asset ratio of the commercial banking system to more directly test the role of banking sector difficulties.
We consider whether the increased popularity of stock and bond funds has contributed to the weakness in the broader monetary aggregates by including the five-year
Treasury security rate as a proxy for interest rates on these alternative assets. We also
consider whether the spread between the five-year rate and a short-term rate, the threemonth Treasury bill rate, is significant. Finally, we consider whether the scaling back
of household borrowing had an impact on the demand for money, by including the rate
on automobile loans as a representative consumer loan rate.20
Each of the loan and interest rate variables is added one at a time to the money demand equation. The equation is reestimated over the entire time period, first with each
variable included over the entire sample period, and then with the variable beginning in
1982-1V. These two specifications allow us to consider whether each variable only became significant, or became increasingly significant, during the 1980s.21
As Table 4 indicates, all of the credit aggregates are significant over both time periods, with their significance increasing during the 1980s.22 The positive coefficients indicate that credit growth and money growth move in the same direction. As growth in
all of these credit measures has been weaker than expected over the last few years, these
20
The interest rates we consider arc similar to the rates examined by Fcinman and Porter (1992).
21
Although we might also want to consider the importance of these variables over the 1989-92 period, this
time period is too short to generate statistically meaningful parameter estimates.
22
The addition of each variable alters the coefficients of the remaining variables slightly, but the changes are
quite small. Hence, we report only the coefficients and standard errors of each additional variable
Table 4: Importance of Credit Variables in M2 Demand Equation
416
Additional Explanatory Variable
1964-IV to 1992-111
1982-1V to 1992-111
Total Dl credit
.20
(.04) **
.25
(.07) ***
Bank credit
.10
(.03) ***
.19
(.08) **
Dl loans
.12
(.04) ***
.24
(.06) ***
.08
(.04) **
.27
(.07) ***
.04
(.04)
-1.38
(.41)***
Five year Treasury rate
-.14
(.09)
-.02
(.05)
Five year minus 3 month Treasury rate
.08
(.06)
-.10
(.10)
-.50
(.16)***
-.14
(.28)
Bank loans
Capital asset
a
Auto loan rate
b
Notes: Standard errors are in parentheses.
a
b
The capital-asset ratio is only available beginning in 1980-1.
The auto loan rate is only available beginning in 1972-1.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
findings are consistent with weak credit growth slowing the growth in M2. 23
Bank loan growth is the weakest over the entire period, having both the smallest coefficient and the lowest significance level (5 percent). But the bank loan variable has
the highest coefficient during the 1980s. Somewhat surprisingly, the capital-asset ratio
is negatively related to the demand for M2 during the 1980s. But this finding can be
explained by the fact that this variable has been rising for the banking sector as a whole,
despite regional banking problems where this ratio has fallen. This finding suggests that
a more disaggregate analysis is likely needed to uncover a link between deposit growth
and measures reflecting the strength of banks' balance sheets.
The table also indicates that although the coefficients generally have the expected
negative sign, the interest rates are for the most part not significant. Their insignificance
likely reflects the fact that only over the last few years have interest rates on alternative
financial assets strongly affected the demand for M2. This recent time period is too
short to obtain meaningful statistics for these variables.
We conducted one further exercise to obtain a better sense of how much of the weakness in the demand for M2 can be accounted for by these credit variables. We examined
the cumulative errors in the predicted values of M2 estimated with and without both a
credit variable (bank loans) and an interest rate variable (the five-year Treasury bond
rate). The results appear in Chart 11. As the chart shows, the original Board of Governors model has a cumulative error in predicting M2 of roughly $250 billion by the third
quarter of 1992. This amount is similar to the measures of "unusual" weakness in M2
reported in the previous section. When the additional variables are added this error falls
to roughly $25 billion.
23
Loan variables were also found to be significant in non-M2 M3 and term deposit demand equations in
Motley (1988).
Chart 11: Cumulative Errors from M2 Demand Equations
Billions of dollars
100
Alternative Model
-200
-250
Li
1987
88
89
90
91
92
Note: Alternative Model includes growth in commercial bank loans, and the five-year
bond rate.
417
Causes and Consequences
To summarize this section, both the forecasting exercise and the plots of residuals
indicate that the behavior of M2 changed after 1989. Additional work suggests that the
weakness in credit growth could account for much of the weak M2 growth. However,
this last conclusion must be viewed tentatively. The improvement in the money demand residuals obtained by including various credit proxies largely results from the
high correlation between credit and money growth, which need not imply causation. In
the next sections, we use alternative approaches to determine whether and to what degree the recent slowdown in the broad monetary aggregates can be attributed to the 1990
credit slowdown.
Causality Tests
418
We now consider whether or not credit growth "Granger causes" money growth. Such
a finding would be consistent with the view that the recent slowdown in credit growth
"led" to the slowdown in the monetary aggregates. However, such a result would not
necessarily mean that the credit growth slowdown caused the money growth slowdown:
a third factor could be responsible for both. Nevertheless, to better understand the importance of the credit growth slowdown for the weakness in the monetary aggregates,
we consider whether a causal relationship exists between the two measures, and whether
the relationship may have changed in recent years.
To examine whether lagged credit measures can "explain" movements in the monetary aggregates, we look at quarterly data from 1959-1 to 1992-1. We first examine
whether the two variables were linked prior to the deregulation of the 1980s. We then
examine their relationship during the 1980s. We also examine whether money growth
"causes" credit growth because the possibility that causality is bidirectional can not be
ruled out. Therefore, we also consider whether lagged values of growth in the monetary
aggregates are significant in explaining credit growth.
Four credit aggregates are used in our analysis, total credit and total lending, for both
commercial banks and for all depository institutions. The two measures of money used,
M2 and M3, include either deposits held at the commercial banks or at all financial institutions, depending upon whether a bank credit or a total depository credit measure
was used in the regression. Including and then excluding the thrift institutions allows
us to investigate whether the link between the credit and money slowdown has operated
through the thrift institutions. This breakdown also allows us to consider whether the
large number of recent thrift failures distorted the money-credit relationship.
Because the series are nonstationary in their level form, the data are all transformed
to growth rates. The same number of lagged values of the independent and dependent
variables are included as explanatory variables. A test for the joint significance of the
lags of the independent variable is then calculated to determine whether movements in
this variable are significant in explaining movements in the dependent variable, after
lagged values of the dependent variable have been taken into account. The tests are performed for each credit variable relative to each money measure, and then the causal link
from money to credit is considered. The marginal significance level of each test is reported in Table 5.
As the table indicates, we conduct the tests using lag-lengths of four and eight quarters. The top half of the table presents the credit variables' predictive power for the monetary aggregates. The results indicate that, generally speaking, the link between the
bank credit variables and the monetary aggregates strengthened during the 1980s, while
the link between lending at all institutions and the similarly measured monetary aggre-
Table 5: Time Series Analysis
A. Bivariate Tests of the Predictive Power of Depository Credit
Independent Variables
Marginal Significance Levels
Lag-Length
TC
TBC
TL
TBL
M2
4
8
.70
.36
.09
.14
.00
.00
.00
.01
M3a
4
8
.83
.94
.03
.31
CO CO
.34
.61
M2
4
8
.59
.87
.00
.01
.71
.72
.00
.00
M3
4
8
.50
.13
.05
.09
.88
.43
.04
.30
1959-1 to 1982-1V
1982-1V to 1992-1
B. Bivariate Tests of the Predictive Power of Money
Dependent Variables
Marginal Significance Levels
Lag-Length
TC
TBC
TL
TBL
M2
4
8
.07
.04
.55
.35
.01
.03
.20
.32
M3a
4
8
.00
.00
.80
.19
.05
.06
.41
.29
M2
4
8
.25
.75
.47
.62
.50
.30
.26
.37
M3
4
8
.16
.23
.98
.92
.27
.41
.33
.75
1959-1 to 1982-1V
1982-1V to 1992-1
Notes: TC, TL = total credit and total loans at all domestically chartered depository
institutions, respectively. TBC, TBL = total credit and total loans at all domestically chartered
commercial banks. M2 = the M2 monetary aggregate, M3 = the M3 money aggregate.
For the tests with TC and TL, the monetary aggregates include deposits at all depository
institutions. For the tests with TBC and TBL, the monetary aggregates include deposits at
all commercial banks.
a. The tests involving M3 at commercial banks begin in 1970-1.
419
Causes and Consequences
gates weakened.24 In the earlier part of the sample period, total bank credit (TBC) and
both loan variables (TL and TBL) are significant in predicting M2. Total loans lose their
significance in predicting M2 in the 1980s, but the link between bank credit and M2 increases. Bank credit is also significant in predicting M3 over both time periods, and the
link between bank loans and M3 becomes significant during the 1980s.
The bottom half of the table presents the results for the causal link running from the
monetary aggregates to the credit variables. The tests indicate that the aggregates are
significant in explaining total credit (TC) and loans (TL) of all financial institutions over
the earlier period, but not during the 1980s. Thus, prior to 1983 there was a bidirectional
relationship only between M2 and total loans. Since that time there has been a unidirectional relationship running from bank credit and bank loans to the monetary aggregates.
Overall, the results seem to support the hypothesis suggested by the earlier discussion, that the link running from bank credit to the monetary aggregates increased following deposit deregulation. This finding is consistent with the view that banks
increasingly made use of managed liabilities to fund their desired levels of credit extensions. This finding also suggests that the slowdown in money growth could have followed from the slowdown in credit growth, and that such a directional link is more likely
to have occurred in the recent credit slowdown than in prior episodes. However, this
conclusion generally holds for the bank credit measures, and not the total depository
credit measures.
Cross-Section Analysis
420
We now attempt to determine more precisely how much of the weakness in the monetary
aggregates can be explained by the recent problems in the banking sector that have constrained credit supply by conducting a cross-sectional analysis of deposit data 25 Because of the regional nature of many of the banking difficulties, more aggregated data
obscures the links between bank balance sheet problems and deposit growth. We begin
by providing an overall picture of deposit growth according to regions of the country.
We then present a more detailed analysis.
Deposit growth by region
Table 6 presents deposit growth by region for the four main categories of deposits—
checking, savings, small time deposits, and large time deposits—at all domestically
chartered U.S. banks. Growth rates for each of these categories and for their total are
shown for 1987-89 and 1989-91 for the nine census regions in order to examine regional
deposit growth both prior to and during the deposit growth slowdown. As the table
shows, growth in total deposits was generally much stronger during the earlier time period than in the more recent period, although there is some variation across types of deposits. The differences in growth rates across the two time periods were relatively small
for checkable deposits: in some cases, growth in checkable deposits picked up and in
some cases it slowed. Growth in total savings accounts rose in every region in the country, while growth in both small and large time deposits plummeted.
The regions of the country which had the most banking sector difficulties also experienced the largest slowdown in deposit growth. For example, New England and the
Mid-Atlantic regions recorded the largest drops in total deposit growth. These two re24
The results were similar when we split the sample at 1979-IH.
25
This section parallels the cross-sectional work found in Lown and Wenninger.
gions had the smallest increases in savings deposit growth, the largest declines in small
time deposits and, with the exception of the East North Central region, the largest declines in large time deposit growth as well. As discussed in Bernanke and Lown and in
Lown and Wenninger, New England and the Mid-Atlantic states experienced the largest
slowdown in lending growth as well. Thus, the table clearly suggests that the slowdown
in credit growth played a role in the deposit growth slowdown.
Table 6: Commercial Bank Deposits by Region
Annualized Fourth Quarter to Fourth Quarter Growth Rates
s.
w.s.
New
England
MidAtlantic
E.N.
Central
W.N.
Central
Atlantic
E.S.
Central
Central
Mountain
Pacific
1987-89
4.5
1.7
4.8
4.2
4.8
3.7
11.5
3.2
9.3
1989-92
8.7
4.3
9.9
9.5
12.0
11.8
10.8
11.8
12.1
1987-89
6.0
4.3
-1.7
0.5
1.2
3.6
4.1
-1.6
4.5
1989-92
15.5
12.9
13.8
13.1
13.2
16.0
9.5
14.2
19.9
1987-89
28.0
21.6
15.5
15.0
22.3
14.9
17.2
13.1
18.6
1989-92
-2.0
0.2
2.3
3.8
2.3
1.9
1.9
-2.6
4.1
1987-89
16.7
12.1
23.4
3.8
20.4
10.6
1.9
2.1
20.7
1989-92
-25.0
-24.2
-11.4
-15.3
-11.9
•6.8
-10.0
-15.2
-10.6
1987-89
12.8
8.4
8.8
7.4
10.9
8.9
9.6
4.2
11.3
1989-92
2.7
1.1
5.4
5.3
5.9
6.3
4.2
5.8
9.7
1987-89
0.03
0.22
0.17
0.10
0.16
0.06
0.09
0.05
0.13
1989-92
0.03
0.20
0.17
0.10
0.17
0.06
0.09
0.05
0.14
Total checking
Total savings
Total small
Total large
Total deposits
Percentage of deposits as a national total
New England = Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont
Mid-Atlantic = New Jersey, New York, Pennsylvania
East North Central = Illinois, Indiana, Michigan, Ohio, Wisconsin
West North Central = Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota,
South Dakota
South Atlantic = Delaware, District of Columbia, Florida, Georgia, Maryland,
North Carolina, South Carolina, Virginia, West Virginia
East South Central = Alabama, Kentucky, Mississippi, Tennessee
West South Central = Arkansas, Louisiana, Oklahoma, Texas
Mountain = Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming
Pacific = Alaska, California, Hawaii, Oregon, Washington
421
Causes and Consequences
422
State- and bank-level regressions
To better determine how important banking sector difficulties were in the money growth
slowdown, we estimate cross-sectional regressions to examine the link between proxies
for banking sector difficulties and deposit growth. We first estimate the regressions using state-level data and subsequently make use of bank-level data. We examine the link
between growth in all four categories of deposits and three different measures of bank
balance sheet health. Deposit changes are measured over three years, 1988-89, 198990 and 1990-91, allowing us to consider whether growth in the monetary aggregates became increasingly tied to the health of banks as financial sector conditions deteriorated.
The three bank balance sheet measures we consider are the capital-asset ratio, loan-loss
reserves as a percentage of total loans, and lagged total loan growth. This last variable
allows us to test for a direct link between credit growth and money growth. The other
two variables focus explicitly on factors that affect the willingness or ability of banks to
supply loans and that thereby could have indirectly affected deposit growth. We also
include in all the regressions employment growth by state to control for the impact of
economic activity on money growth.26
Our expectations are that bank balance sheet measures and lagged loan growth may
matter less for the 1988-89 regressions than for the 1990-91 regressions. This asymmetric
expectation is based on the view that a bank's balance sheet might impose a constraint on
credit formation when the bank is in poor financial health, but when the bank is in good
financial health the pace of credit formation might not be closely linked to the same balance sheet measures. We also expect that the balance sheet measures would be more important in explaining small and large time deposits than in explaining checkable deposits
because, as we discussed earlier, banks typically use time deposits as managed liabilities.
There are, however, limitations to the use of state-level data for an analysis of deposit
growth. The asymmetric effect of balance sheet constraints on deposit formation described above could be obscured when bank data is aggregated to the state level. Moreover, deposits are not necessarily limited by state borders, especially in the case of
brokered small and large time deposits. Forcing such an aggregation might produce unexpected or counterintuitive results. Nevertheless, we begin with an analysis of statelevel data and later conduct a similar examination of bank balance sheet health and deposit growth with bank-level data.
State level results
The state-level regression results are reported in Table 7 and indicate that the relations
among the variables are roughly in line with expectations. Checkable deposits are explained only by employment growth, with lagged loan growth and the bank balance
sheet measures having no significant explanatory power. This finding is consistent with
the notion that checkable deposits are largely demand determined. The capital-asset ratio and loan-loss reserves are significant in explaining savings deposits over the 198990 period, although loan losses have a positive rather than a negative sign. Neither variable is significant in the 1990-91 period.
The results for the two time deposit categories are consistent with the notion that
banks treat these accounts as managed liabilities. Time deposits are fairly consistently
explained by lagged loan growth, while the capital-asset ratio also is significant in ex-
26
Lown and Wenninger found a link between these explanatory variables (excluding loan growth) and loan
growth; here we are considering whether the same bank balance sheet variables can explain the weakness
in deposit growth.
plaining large time deposits. Finally, as is the case for savings deposits, small time deposits are unexpectedly positively related to loan losses over the 1989-90 period.
In general, the bank variables are not significant in explaining growth in checkable
deposits, while these variables matter in varying degrees for the other three deposit categories. The capital-asset ratio becomes increasingly significant in explaining large
time deposits, while lagged loan growth fairly consistently explains growth in both
small and large time deposits. These findings are consistent with the observed behavior
of the aggregates which shows the term components of M2 and M3 to be unusually
weak, suggesting that a credit crunch, or a capital crunch, likely played a role in their
sluggish growth.27
27
We also considered lagged loan growth as an instrument for contemporaneous loan growth. These regressions (it quite poorly, indicating that lagged, but not current, loan growth plays a role in explaining deposit
growth.
Table 7: Cross-sectional Regressions
Fourth Quarter to
Fourth Quarter
K/A
Empl.
LL
R2
LG(-1)
1988-89
(1) Checkables
(2) Savings
1
(2 ) Savings
3
1.17 (.24)***
[.33]***
.16 (.45)
[.67]
.82 (.65)
[1.1]
.02 (.06)
[-09]
.42
.28 (.67)
[-65]
-2.26(1.2)*
[1.3]*
.59(1.8)
[1.9]
.30 (.17)*
[.17]*
.08
.25 (.63)
[.66]
-.94(1.3)
[1.0]
-.08(1.7)
[1.4]
.45 (.17)***
[.16]***
.15
.50 (.13)***
[.20]**
.38
2.06 (.25)*** [.34]***
.67
(3) Small time
-.18 (.50)
[.76]
-1.01 (.94)
[1.7]
-1.58(1.4)
[f2.4]
(4) Large time
2.59 (.99)***
[1.0]**
-1.70(1.8)
[1.4]
4.03 (2.7)
[2.4]
(1) Checkables
1.37 (.22)***
[.18]***
.13 (.53)
[•61]
-.08 (.58)
[.70]
(2) Savings
1989-90
.02 (.07)
[08]
.46
1.47 (.45)***
[.52]***
1.92(1.1)*
[.83]**
3.92(1.2)*** [1.2]***
.17 (.14)
[-15]
.27
(3) Small time
.72 (.55)
[.71]
.21(1.4)
[1.3]
4.07(1.5)*** [2.3]**
-.03 (.17)
[.18]
.15
(4) Large time
-.25 (.75)
[.84]
3.74(1.8)**
[2.2]*
.86 (2.0)
[2.6]
.26 (.40)
[.48]
1.02 (.60)*
[-63]
-.52 (.94)
[1.3]
-.09(1.3)
1.00 (.24)*** [.21]***
.29
1990-91
(1) Checkables
b
[.14]
.08
.69 (.32)*
[.32]**
.80 (.54)
[.65]
-.25 (.84)
[1.1]
-.01 (.12)
[.11]
.20
(2) Savings
.34 (.69)
[1.1]
1.17(1.0)
[1.4]
-.69(1.6)
[2.0]
.02 (.23)
[-29]
.04
(3) Small time
-.16 (.86)
[.77]
2.20(1.3)*
[2.8]
.83(2.1)
[2.4]
.56 (.29)*
[.27]**
.19
3.6(1.2)***
[1.4]***
1.18(1.9)
[1.6]
.51 (.27)*
[.27]*
.36
(1')Checkables
(4) Large time
.48 (.81)
[.79]
Notes: Dependent variable is the fourth quarter to fourth quarter growth rate of the specified category of deposits aggregated to the state level. Empl. is the December to December growth rate of unemployment. K/A is equity capital relative
to total assets at the beginning of the period. LL is the level of loan loss reserves relative to total loans at the beginning
of the period. LG(-1) is total loan growth fourth quarter to fourth quarter lagged one year. Conventional OLS standard
errors are in parenthesis. Standard errors based on White's correction for heteroskedasticity are in brackets.
a
- This regression excludes the outlier Texas, which experienced large growth in savings accounts, despite a low
K/A ratio and exerted a large influence on the K/A coefficient.
b
- This regression excludes Rhode Island which experienced huge growth in checkables despite a large decline in
unemployment.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
423
Causes and Consequences
The one unexpected result is the positive and significant coefficient on loan-loss reserves in explaining savings and small time deposit growth during 1989-90. A possible
explanation for this finding is that banks with large loan losses moved away from issuing
large, uninsured time deposits towards insured savings and small time deposits during
this period. But a negative relationship between large time deposits and loan losses
should then have emerged, and did not. Alternatively, loan-loss reserves could be serving as a proxy for some omitted variable. For example, banks that moved away from
loan issuance and into security holdings in the face of large loan losses might have continued to issue savings deposits. Loan-loss reserves would be positively correlated with
growth in security holdings and, because we did not include security holdings in the regressions, loan losses could be capturing this relationship.
Bank level results
We next examined the link between bank balance sheet variables and deposit growth at
the bank level. We again examined each of the four categories of deposit growth, but
only over the 1990-91 period. In this case, we are assuming that all banks behave identically, but we are not forcing deposits to be limited by state boundaries.
The results from the bank-level regressions, reported in Table 8, indicate that the capital-asset variable is highly significant in explaining each of the deposit categories. The
loan-loss variable is significant only in the checkables equation and has the expected
negative sign. In no case is lagged loan growth significant (These results are not reported.) Thus, the significant relationship between loan losses and deposit growth found in
the state-level analysis is not found using bank level data, although the coefficient on
loan losses is still positive for both savings and small time deposits. Moreover, the loanloss coefficient is negative for large time deposits. While the insignificance of these
coefficients makes one hesitant to attach much weight to them, they do suggest that
banks moved away from large time deposits and towards savings and small time deposits in response to weakening balance sheets.
Table 8: Cross-sectional Bank Regressions
1990-1V to 1991-IV; 10,700 Observations
424
Dependent
Variable
Const.
K/A
LL
R2
(1) Checkables
-0.18 (.12)
4.96 (.13)***
1.42(0.42)***
.13
(2) Savings
-0.14 (.67)
5.97 (.73)***
3.59 (2.26)
.05
(3) Small time
-0.52 (.60)
8.53 (.65)***
1.14(2.01)
.03
(4) Large time
-0.78 (.14)***
5.97 (.16)***
-0.63 (0.49)
.13
Notes: Dependent variable is the fourth quarter to fourth quarter growth rate of the specified
category of bank deposits. K/A is equity capital relative to total assets at the beginning of the
period. LL is the level of loan loss reserves relative to total loans at the beginning of the period.
OLS standard errors are in parenthesis. Dummy variables by state were also included in the
regressions to control for region specific variation.
*** Significant at the 1 percent level.
Explaining slow money growth
Finally, we use the bank-level regression estimates to determine how much of the slowdown in money growth can be attributed to the bank balance sheet variables. Using the
1990-IV to 1991-IV bank level regression results, we calculate how much of the slowdown in time deposit growth can be explained by the bank balance sheet variables. To
do so, we determine the changes in the regional capital-asset ratios and loan-loss variables from 1988-1V to 1990-IV when the banking system experienced the most distress.
We then compute to what degree these changes can account for the slowdown in regional deposit growth from the period 1988-1V through 1989-1V to 1990-IV through 1991IV.
Large time deposits in the New England and the Mid-Atlantic regions grew 16 percent and 8 percent, respectively, from 1988-IV to 1989-IV, and declined 27 percent over
the 1990-IV to 1991-IV period in both regions. Hence these regions recorded net
swings in large time deposit growth rates of negative 43 and 35 percentage points, respectively. Changes in the balance sheet variables account for 6.0 percentage points and
3.3 percentage points of these declines, or 14 and 10 percent of the total decline in large
time deposit growth rates in New England and the Mid-Atlantic regions, respectively.
A similar calculation finds that, over the same time period, small time deposit growth
fell 34 and 24 percentage points in these two regions. The balance sheet variables account for roughly 8 percent of these declines (2.8 and 2.0 percentage points, respectively). In the remaining regions, which experienced much smaller deteriorations in their
balance sheet variables, the balance sheet variables explain a smaller amount of the declines in time deposits or even serve to partly offset declines.
The low explanatory power of the balance sheet variables must be viewed with caution, however, because of the drawbacks of this exercise. We assumed that the coefficients on the independent variables estimated across banks could be used in a rough way
to explain the deposit slowdown over time in each region. This required moving from
cross-sectional estimates to time series data, and from bank-level to regional data. More
to the point, regional growth in small and large time deposits fell by large amounts over
the last few years, but the regional deterioration in capital-asset ratios and increases in
loan losses are not big enough to account for such large declines. Presumably, if we
could conduct this exercise using bank-level balance sheet data, we could account for
the large drops in time deposits. But, even still, the regression coefficients are based on
the unrealistic assumption that all banks behave identically. Hence, even if a bank-level
accounting exercise were feasible, it still might not be able to explain the large declines
in deposit growth.28
Overall, the data indicate that the regions experiencing the most severe financial sector difficulties also experienced the largest declines in deposit growth. The regression
analysis indicates that the bank balance sheet variables, and in particular the capital-asset ratios, are highly significant in explaining weak deposit growth. But calculations using regional data do not indicate that these variables can account for much of the
weakness in regional deposit growth. This latter finding most likely results from the assumption behind the regression analysis that all banks behave identically and from the
use of these bank-level regressions to interpret regional deposit growth.
28
See Boyd and Gertler (1993) for a discussion of how bank behavior differs by asset size.
425
Causes and Consequences
IV. Conclusion
426
In this study, we have assessed the degree to which the unusual weakness in the broad
monetary aggregates, particularly M2, over the past few years has been associated with
factors that have constrained credit formation. Using a variety of methods, we find that
factors that have shifted credit demand and supply and money demand—apart from direct business cycle influences associated with the 1990-91 economic downturn—account for a considerable portion of the recent weakness in M2. A decreased willingness
on the part of depository institutions to lend, retrenchment in loan demand stemming
from efforts of businesses and households to reduce their debt exposure, and shifts in
consumer preferences for holding depository liabilities are among the factors that have
simultaneously weakened bank credit formation and deposit creation. Further work also
suggests that changes in bank balance sheet behavior over the past decade may have increased the impact of the credit slowdown on the M2 aggregate. Overall, the totality of
the evidence found in this study supports the view that the slowdown in lending at depository institutions played a significant role in weakening the broad monetary aggregates. Moreover, while the effects of supply-side credit constraints cannot be fully
separated from other factors, the available evidence points to these supply-side forces as
playing a significant role in slowing credit formation and weakening monetary growth.
As a first approximation, we simply compare the recent behavior of M2 and depository credit measured relative to nominal income since the onset of the 1990-91 recession
to their behavior around past business cycles. These comparisons suggest that by the
middle of 1992 the level of M2 may have been reduced by around 8 to 9 percent as a
result of noncyclical developments that have reduced depository credit formation. This
methodology makes no attempt to discriminate between the different possible causes of
credit or money restraint.
Alternatively, an evaluation of the predictive performance of a conventional econometric equation of M2 also indicates that by the middle of 1992, M2 was about 10 percent below levels that would have been expected, given the actual behavior of the
equation's explanatory variables during the past few years. The prediction errors of the
M2 model reflect a combination of deviations in money demand and disturbances linked
to shifts in credit demand and supply. Additional econometric work also shows that the
predictive performance of the M2 demand equation improves when direct measures of
credit or other factors that capture cutbacks in lending are included.
Econometric tests were performed to examine how the channels between bank credit
and deposit creation may have changed in recent years. The results from Granger causality tests indicate that the links running from depository credit formation to money
creation have strengthened amid the financial innovations of the past decade. This finding is consistent with the observation that banks have come to rely more heavily on
many nontransactions deposits included in the M2 aggregate to fund their credit formation and suggests that M2 has become more sensitive to developments affecting bank
credit formation.
Finally, using cross-section data, we find that direct measures of bank balance sheet
distress, such as capital-asset ratios, can account for at least a modest portion of the observed weakness in several components of the broader monetary aggregates. A deterioration in the financial well-being of the depository sector is felt to have impeded
deposit growth by making banks less willing to lend. Deposits used as managed liabilities have been most affected. Direct measures of loan growth included in these regressions are also found to help explain slow deposit growth.
References
Bernanke, Ben S., and Alan S. Blinder. "Credit, Money, and Aggregate Demand,"
American Economic Review, May 1988.
Bernanke, Ben S., and Cara S. Lown. "The Credit Crunch," Brookings Papers on Eco-
nomic Activity, 2:1991.
Boyd, John, and Mark Gertler. "U.S. Commercial Banking: Trends, Cycles, and Policy." Paper prepared for the 1993 National Bureau of Economic Research Macroeconomics Annual.
Carlson, John B., and Sharon E. Parrott. "The Demand for M2, Opportunity Cost, and
Financial Change," Federal Reserve Bank of Cleveland, Economic Review, Second
Quarter, 1991.
Carlson, John B., and Susan M. Byrne. "Recent Behavior of Velocity: Alternative Measures of Money," Federal Reserve Bank of Cleveland, Economic Review, First Quarter, 1992.
Duca, John V. "The Case of the Missing M2," Federal Reserve Bank of Dallas, Economic Review, Second Quarter 1992.
Feinman, Joshua, and Richard D. Porter. "The Continuing Weakness in M2," Finance
Economics Discussion Series no. 209, Board of Governors, September 1992.
Friedman, Benjamin M., and Kenneth N. Kuttner. "Money, Income, Prices, and Interest
Rates," American Economic Review, June 1992.
Furlong, Frederick., "Capital Regulation and Bank Lending," Federal Reserve Bank of
San Francisco, Economic Review, no. 3, 1992.
Gunther, Jeffrey W., and Kenneth J. Robinson. "The Texas Credit Crunch: Fact or Fiction?" Federal Reserve Bank of Dallas, Financial Industry Studies, June 1991.
Higgins, Bryon. "Policy Implications of Recent M2 Behavior," Federal Reserve Bank
of Kansas City, Economic Review, Third Quarter, 1992.
Kahn, George A. "Does More Money Mean More Bank Loans?" Federal Reserve Bank
of Kansas City, Economic Review, July/August 1991.
Keeton, William R. "Deposit Deregulation, Credit Availability, and Monetary Policy,"
Federal Reserve Bank of Kansas City, Economic Review, June 1986.
Lown, Cara S. "The Credit-Output Link vs. the Money-Output Link: New Evidence,"
Federal Reserve Bank of Dallas, Economic Review, November 1988.
Lown, Cara S., and John Wenninger. "The Role of the Banking System in the Credit
Crunch," this volume.
Mehra, Yash P. "An Error-Correction Model of U.S. M2 Demand," Federal Reserve
Bank of Richmond, Economic Review, May/June 1991.
427
Causes and Consequences
428
Morgan, Donald P. "Are Bank Loans a Force in Monetary Policy?" Federal Reserve
Bank of Kansas City, Economic Review, Second Quarter 1992.
Motley, Brian. "Should M2 Be Redefined?" Federal Reserve Bank of San Francisco,
Economic Review, Winter 1988.
Peek, Joe, and Eric Rosengren. "The Capital Crunch in New England," Federal Reserve
Bank of Boston, New England Economic Review, May/June 1992.
Wenninger, John, and John Partlan. "Small Time Deposits and the Recent Weakness in
M2," Federal Reserve Bank of New York, Quarterly Review, Spring 1992.
The Credit Slowdown Abroad
by S. Hickok and C. Osier*
Introduction
Over the last few years credit growth has slowed sharply in most industrialized countries. This paper examines the factors behind this credit slowdown. We concentrate on
the recent behavior of credit in Japan, the United Kingdom, and France. These countries are chosen because of their prominence and the broad range of their credit experiences. We put these countries' experiences into an historical perspective and then
undertake an econ-ometric examination of the factors lying behind them. In addition,
we discuss the similarities among credit developments in these and other major foreign
industrial economies.
We find that the slowdowns in credit growth in Japan, the United Kingdom, and
France reflect in large part the return of credit growth rates to more normal levels after
exceptionally rapid growth in the 1980s. Cyclical factors—the rapid rise and then slowdown in GDP growth—played a role in these credit dynamics, particularly in France,
but noncyclical factors were generally more important. The financial market deregulation and innovations that took place in most industrial countries during the 1980s were
the most significant noncyclical factors. Deregulation, most notably the end of credit
controls, was generally followed by a period of fast growth in credit as previously rationed sectors gained improved access to credit markets. This source of growth would
naturally have tapered off as such agents, typically consumers, began to reach their desired borrowing positions. We find that this deregulation-induced swing in credit growth
was most important in France, where deregulation was most recent, but was important
elsewhere as well. Other financial market developments, such as the introduction of
commercial paper markets in all three countries and the sharply increased issuance of
equity-linked bonds in Japan, led to temporary surges and declines in individual categories of credit as agents shifted their portfolios to take advantage of new opportunities.
More generally, financial innovation led to a deepening in financial intermediation and
permanently higher credit/GDP ratios.
1
Our thanks to R.G. Davis, M.A. Akhtar, C. Pigott, A. Rodrigues, and R. Seth for comments. Michael
Hansen provided extensive assistance with this paper.
429
Causes and Consequences
Financial market deregulation and innovations were primary forces behind rapidly
rising equity and real estate prices in the 1980s, and these asset price dynamics in turn
contributed significantly to the surge and subsequent slowdown in credit growth.
Agents' wealth and their ability to borrow increased along with asset prices, as did the
amount agents needed to borrow to purchase any individual asset, providing impetus to
credit growth. Asset prices eventually began to slow, which reduced growth in agents'
borrowing capacity and lending. This wealth-related effect of asset prices on credit
seems to have been important in all three countries. Fast asset price increases also attracted speculative borrowing during the late 1980s, and then declining prices discouraged such borrowing. This relatively transitory speculative element is estimated to have
been most important in shaping credit developments in Japan and least important in
France.
The recent slowdown in credit growth abroad was also influenced in some countries
by increased attention to capital adequacy levels in the late 1980s and early 1990s, in
response to the announcement by the Bank for International Settlements (BIS) of new
bank capital standards. This source appears to have contributed significantly to the
credit slowdown in Japan but not to the credit slowdowns in the United Kingdom and
France.
The pattern we observe in our three focus countries appears in most other industrialized economies. Germany, Italy, Spain, Sweden, Switzerland, and Australia generally
experienced rising credit growth which peaked at levels much higher than GDP growth
and subsided thereafter. As in our focus countries the rise and fall in credit growth generally followed deregulation and other financial market changes and was accompanied
by rapidly rising and then falling asset prices, particularly in real estate. Credit adjustment problems tied to speculative asset market dynamics, in particular, appear to have
been substantial in many countries where credit surged markedly. Germany stands out
as having little deregulation, no burst of credit growth, and no major developments in
asset prices; Sweden stands out as being relatively extreme in all these dimensions.
Section I of our study provides an historical perspective on credit market developments in Japan, the United Kingdom, and France. In Section II we support these narrative observations with econometric analysis of credit behavior in each country. Section
III summarizes credit developments in the other foreign countries and in Section IV we
conclude.
I: Historical Perspective on Credit Developments in Japan, the United Kingdom, and France
430
Nominal credit growth slowed markedly after 1989 in Japan, the United Kingdom, and
France (Charts 1-3), and by 1991 it had generally fallen to half or less than half of the
rates of growth registered during the late 1980s. The U.K. credit slowdown was particularly dramatic: bank credit grew only 2 1/2 percent in 1991, down from 30 percent in
1988. The credit slowdown in each of the countries was widespread as well as severe,
affecting most types of loans and most categories of borrowers and lenders.
In this section we present these credit developments in an historical perspective. For
each of the three focus countries we provide first an overview of credit developments
since 1980 and then an in-depth discussion of the factors behind those developments,
particularly the recent credit slowdown. We highlight the contributions of financial
market deregulation and innovation, asset market developments, cyclical factors, and
credit supply constraints.
Chart 1: Credit and GDP: Japan
Trillions of Yen
1000
Total Credit
800
600
Percent
25
Four Quarter Growth Rates
Total Credit
Bank Credit
Nominal GDP
Percent
800
Credit as a Share of GDP
Total Credit
700
600
500
400
Bank Credit
300
200
U
1974
76
78
80
82
84
86
88
90 91
431
Causes and Consequences
Chart 2: Credit and GDP: United Kingdom
432
Billions of Pounds
700
Total Credit
Percent
60
Four Quarter Growth Rates
Bank Credit
Total Credit
>»«/ \
i
-10
Percent
500
Credit as a Share of GDP
Total Credit
400
300
200
••-**
Bank Credit
100
0
u
1970
72
74
76
78
80
82
84
86
88
90 91
Chart 3: Credit and GDP: France
Trillions of Francs
7
Total Credit
Percent
25
Four Quarter Growth Rates
Bank Credit
20
15
10
Nominal GDP
Percent
400
Credit as a Share of GDP
350
Total Credit
300
250
200
Bank Credit
150
100
50
1974
76
78
80
82
84
86
88
90 91
433
Causes and Consequences
We focus on credit to the private nonfinancial sector and use two measures of credit,
"bank credit" and "total credit." "Bank credit," which includes loans to commercial and
industrial firms, consumer loans, and mortgages, is of primary interest to those concerned with the banking sector per se. "Total credit," which includes bonds issued by
nonfinancial companies, commercial paper, and mortgages issued by non-bank financial
firms, as well as bank credit itself, is of greatest relevance to those interested in economic
growth.
Japan
As shown in Chart 1, credit growth in Japan slowed dramatically in 1991. This slowdown must be evaluated in light of the extremely rapid growth which preceded it, also
shown in Chart 1. A brief historical and cyclical perspective on Japanese credit will be
helpful before we turn to analyze the underlying determinants of recent credit developments.
Table 1 presents a disaggregated view of recent Japanese credit developments.
Growth in credit and GDP were fairly comparable during the 1970s but diverged sharply
thereafter. Despite the slowdown in nominal GDP growth in the early 1980s, growth in
Table 1: Japan: Outstanding Growth in Private Nonfinancial Credit
Average Annual Growth Rate
434
1974-1V 1977-1V 1981-IV 1982-1V 1985-1V 1986-1V 1989-1V 1990-1V
to
to
to
to
to
to
to
to
1977-1V 1981-IV 1982-1V 1985-1V 1986-1V 1989-1V 1990-1V 1991-IV
Nominal GDP
10.7
8.2
4.2
6.2
3.6
6.3
7.1
5.1
Real GDP
4.4
4.2
3.9
4.2
2.5
5.3
4.7
3.2
CPI
8.0
5.1
2.2
1.8
-0.7
1.5
3.3
2.8
Total credit
12.2
9.4
8.8
9.8
8.9
13.5
11.5
5.9
Bank credit of
which to:
n.a.
7.6
9.0
7.9
7.8
7.8
7.8
4.1
Corporations
n.a.
7.0
9.5
8.3
7.1
5.6
6.4
3.5
Consumers and
unincorporated
businesses
n.a.
7.1
8.9
16.0
25.8
30.1
22.8
7.2
13.4
4.9
2.9
9.3
17.1
10.2
6.0
Mortgages
Nonbank credit
of which:
n.a
11.7
8.6
11.8
10.0
18.6
14.3
7.2
Bonds
14.1
9.0
5.6
29.1 a
13.9
26.9
18.3
11.9
Commercial
paper
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
24.7
-22.9
a
- Foreign currency bonds are included in bond values outstanding from 1985-1V on; growth in
domestic bonds alone averaged 15.7 percent over the 1982-1V to 1985-1V period.
almost all types of credit was quite robust.2 Credit to consumers and small (unincorporated) businesses rose particularly rapidly.
The 1986 economic slowdown in Japan had no apparent effect on the rate of credit
growth. Mortgage credit growth increased and credit to consumers and small businesses
accelerated even further. The sharp divergence of GDP and credit growth continued
during the rapid economic growth period of 1987-89. Nonbank credit growth surged to
a 19 percent annual rate, triple the rate of GDP growth, an acceleration which was
spurred in part by the introduction of a commercial paper market. Bank credit growth,
on the other hand, remained high during the recovery but did not accelerate. The banks'
traditional corporate customers increasingly met their credit needs in the nonbank securities market, a force which offset soaring growth of bank lending for mortgages, consumers loans, and small businesses.
Credit growth remained strong through 1990, then abruptly fell off. The deceleration
was pronounced in all types of credit instruments and to all credit borrowers. The sharpest decline, however, was in the commercial paper market, where credit outstanding actually fell by 23 percent. The slowdown in bank credit growth to consumers and small
business was also acute.
Viewed over the entire 1980s, Japanese credit growth clearly exhibited a sharp rise
relative to economic activity (GDP). This rise was especially evident in 1982 and 1986
when credit growth picked up and then maintained its momentum despite slowdowns in
the Japanese economy. Only in 1991 did the rising trend in Japanese credit growth appear to have waned. We will now investigate the primary factors—financial deregulation coupled with financial innovation, asset price developments, and monetary and
capital adequacy policies—shaping this credit behavior. During our investigation we
will distinguish between those factors likely to have had a fundamental (that is, sustained) impact on the credit-GDP relationship and those factors that probably only temporarily altered this relationship during the last decade.
Sources of Rapid Credit Growth in the 1980
A key factor leading to a sustained rise in credit relative to GDP was the deregulation
of Japanese credit markets in the early 1980s. Of particular note, Bank of Japan "window guidance" credit controls on banks, occasionally binding in the 1970s, were employed much less restrictively in the 1980s.3 One result of the loosening in credit
controls was that loans to individuals and small business, to a large extent rationed out
in the 1970s, rose sharply in proportion to loans to manufacturing corporations (in Table
1 this can be inferred from the much higher growth rates of loans to individuals and
small businesses than loans to manufacturing corporations). Much of the growth in consumer and small business loans represented a credit stock adjustment as previously unsatisfied credit demand was met and the stock of credit outstanding permanently rose
relative to the level of GDP.
A shift in corporate borrowing to nonbank sources, reflecting both deregulation and
financial innovation, facilitated bank lending to consumers and small businesses. The
introduction of commercial paper in 1987 was a salient example of deregulation and innovation spurring nonbank credit growth. Newly sanctioned Japanese borrowing in the
2
Only mortgage growth slowed along with GDP growth in the early 1980s. However, mortgage credit
had grown very rapidly in the 1970s due to government incentives.
3
OECD Economic Survey on Japan, July 1984, discusses the declining importance of window guidance
as a policy instrument. Sec also Cargill and Royama (1992).
435
Causes and Consequences
Euromarkets was another example. Technological advances in financial services along
with new hedging and other financing options further promoted growth in corporate borrowing, especially outside of traditional "main bank" channels.4 Again, here was a fundamental shift in credit dynamics that affected the credit-GDP relationship during the
1980s.
The mid-1980s increase in the spread of bank lending rates over government bond
yields was, surprisingly, clear evidence of the significant impact of financial market deregulation and liberalization. Financial opening to foreign investors tended to push government bond yields down. At the same time the lifting of interest rate ceilings tended
to raise private borrowing rates.5 In consequence, by the mid-1980s the mortgage
spread over government bonds was well over 100 basis points, compared to a negligible
or negative spread in the early 1980s (Chart 4).
Another important factor promoting, as well as being promoted by, the rise in credit
growth was a rapid rise in stock and real estate prices (Chart 5). 6 Land prices almost
doubled in the 1980s while equity prices increased fivefold. These sharply escalating
prices were in part a result of rapid credit growth but the rising asset prices, in turn,
spurred credit growth even further.7 Specifically, rapidly rising stock prices drastically
4
Japanese corporations have traditionally had a "main bank" that looked after their financial needs.
Ogusi (1990) discusses the role of swaps and other new financing options in diversifying the sources of
corporate funding.
5
Bank of Japan (1990) discusses the impact of interest rate liberalization.
6
Stock price developments shown on Chart 5 are based on index of all share issues listed on the Tokyo
Exchange, published in the International Monetary Fund's International Finance Statistics. Urban land
price developments are based on an index reported in the Nikkei Macro Economic Statistics Data Bank.
7
Japan's strong economic performance relative to other industrial countries also fueled the asset price rise.
Chart 4: Housing vs Government Bonds Yields: Japan
Percentage Points
3
2 -
1 -
1980 81
436
82
83
84
85
86
87
88
89
90
91
reduced the price of corporate convertible and warrant bonds and, hence, led to an explosion in the corporate issuance of these securities in the second half of the 1980s.8 A
sharp drop in bank interest rate spreads over government bonds in the latter 1980s was
in part a result of the emergence of this alternative cheap source of credit. On the real
estate side, rapidly rising land prices provided a growing source of collateral for bank
loans. Soaring asset prices also raised the demand for borrowed funds to support both
speculative and nonspeculative stock and real estate purchases.
These asset price developments clearly had both sustained and temporary effects on
the credit-GDP relationship. Permanently higher asset prices meant permanently higher
credit demand to cover the cost of asset purchases and permanently higher credit supply
in response to permanently higher collateral value. Sharply rising asset prices temporarily raised credit demand for speculative purchases and temporarily raised credit supply in response to the attractiveness of potential gains on warrant and convertible bonds.
Both factors pushed up the credit/GDP ratio in the 1980s.
Monetary easing following the sharp appreciation of the yen in the mid-1980s further
fueled credit growth. The Bank of Japan's discount rate was lowered to a historically
low (in nominal terms) 2 1/2 percent in 1987 while the monetary base grew on average
11 1/2 percent per year between 1986 and 1989, up from 7 percent per year in the preceding four year period. The slowing Japanese economy in 1986 was partly responsible
for
@FedFRASER